Designing human experiences in the age of AI.
Links curated by Peter Polgar.
For a weekly newsletter of hand selected links, signup here.
Posted on 2018 June 4 by polgarp · Tagged in AI design
Original link: https://medium.com/design-ibm/the-designer-of-the-future-51318481650
Cool article series on AI design in three parts: Mindset, methods, modes with even a few useful tools. Well worth the read.
Link to post
Posted on 2018 June 4 by polgarp · Tagged in Google and Military AI
Original link: https://www.nytimes.com/2018/05/30/technology/google-project-maven-pentagon.html
Nice write up on Google’s Pentagon contract. AI tech is still a specialized field, and at least some practitioners recognize the impact they can have on the world through the scale of this tech.
Link to post
Posted on 2018 June 4 by polgarp · Tagged in Introduction and Machine learning
Original link: https://uxdesign.cc/an-intro-to-machine-learning-for-designers-5c74ba100257
As far as intros goes, this post contains a few good points, especially on the what’s possible side. I’m wondering though what you need to know about the different types of machine learning to be able to design products.
Link to post
Posted on 2018 June 3 by polgarp · Tagged in Co-creation, Writing, and Tool
Original link: https://www.theverge.com/2018/5/31/17405042/botnik-studios-predictive-writer-voicebox
Liked this article describing how to use a Voicebox, a writing tool using similar predictive text to your phone keyboard. Maybe this is not true co-creation yet, but you will have the feeling as if you would collaborate with the original owner of the voice. So where the data for the algorithm was from.
Link to post
Posted on 2018 June 3 by polgarp · Tagged in Course
Original link: https://www.elementsofai.com/
Seems like an interesting intro course on AI, suitable especially for non-engineers.
Link to post
Posted on 2018 June 3 by polgarp · Tagged in AI winter and Deep learning
Original link: https://blog.piekniewski.info/2018/05/28/ai-winter-is-well-on-its-way/
Pretty good write up on current trends in AI, that point towards the current AI / deep learning hype to end soon. My biggest conclusion: doing real-world products is much harder then doing experiments.
Link to post
Posted on 2018 May 28 by polgarp · Tagged in AI development
Original link: http://ai.stanford.edu/~zayd/why-is-machine-learning-hard.html
More technical description on the difficulties of AI development. In case you were wondering - compared to issues present at standard software development there will exponentially more issues. So some rules of thumbs need to be re-evaluated.
Link to post
Posted on 2018 May 28 by polgarp · Tagged in Research
Original link: https://www.theverge.com/2018/5/23/17379526/machine-learning-ai-spot-arguments-online-wikipedia
Interesting idea, to have an AI notify conversion participants that they are heading into a too heated situation. Usually when online, things can go sour much faster, as many clues people generally rely on (non-verbal but also contextual) are missing. But maybe augmented with AI we could have better conversations and be more excellent to each others.
Link to post
Posted on 2018 May 28 by polgarp · Tagged in Art and Creativity
Original link: https://medium.com/@naotokui/imaginary-soundscape-take-a-walk-in-soundscapes-imagined-by-ai-f8b99f82eefb
Looks (and sounds) like a nice project. I tried with a few images, and got very accurate results. A good example on how AI can even scale creativity - like editing music for pictures.
Link to post
Posted on 2018 May 28 by polgarp · Tagged in Bias and Overview
Original link: https://thewalrus.ca/how-we-made-ai-as-racist-and-sexist-as-humans/
Nice overview on the history of bias in machine learning. “Garbage in, garbage out” is indeed a good way of describing what some of the companies are building.
Link to post
Posted on 2018 May 28 by polgarp · Tagged in AI design and Design pattern
Original link: https://uxdesign.cc/the-canny-rise-a-ux-pattern-921e400b26ed
Excellent article, describing a new type of interaction design pattern, the canny rise. This patterns helps people understand they are communicating with an AI without sounding too artificial or - too human, skipping the uncanny valley. This is one of the first pattern I see describing an AI-related design pattern and not just iterating on existing UI patterns.
Link to post
Posted on 2018 May 28 by polgarp · Tagged in Bias and Mathwashing
Original link: https://qz.com/1276781/algorithms-are-making-the-same-mistakes-assessing-credit-scores-that-humans-did-a-century-ago/
Interesting article detailing the rise of algorithmic credit scoring. Probably not all systems mentioned are equally bad, but the overall tendency seems to trust the machine - with little regard on the creators own biases, or how they were handled. The difficulty of explaining machine learning decisions seems to be also an issue.
Link to post
Posted on 2018 May 28 by polgarp · Tagged in Resource
Original link: https://medium.com/@karamanbk/39-machine-learning-resources-that-will-help-you-in-every-essential-step-b2696515ed9
Nice list of resources for machine learning. Mostly interesting for engineers, but quite a few good links for people just getting started.
Link to post
Posted on 2018 May 21 by polgarp · Tagged in Research and Interview
Original link: https://www.quantamagazine.org/to-build-truly-intelligent-machines-teach-them-cause-and-effect-20180515/
Very interesting interview with Judea Pearl, who talks about the limits of current AI tech (mostly machine learning), and gives some insights what might come next.
Link to post
Posted on 2018 May 21 by polgarp · Tagged in AI development, Product development, Guide, and Google
Original link: https://developers.google.com/machine-learning/rules-of-ml/
While this guide is more focused on the engineering part, it’s actually an excellent guide for anyone working with AI products.
Link to post
Posted on 2018 May 21 by polgarp · Tagged in AI fail, Police, and Facial recognition
Original link: https://www.independent.co.uk/news/uk/home-news/met-police-facial-recognition-success-south-wales-trial-home-office-false-positive-a8345036.html
Although it’s just an experiment, and it’s just fair to test a new system before deploy, but the false positive rate here is astonishingly bad. The original training set and / or the algorithm must be deeply flawed. I have to wonder how many false negatives does their system produce?
Link to post
Posted on 2018 May 21 by polgarp · Tagged in Fun
Original link: https://www.bloomberg.com/news/features/2018-05-17/i-tried-to-get-an-ai-to-write-this-story-paul-ford
“Then I gave up and wrote it my damn self.” Recommended for anyone trying to apply machine learning.
Link to post
Posted on 2018 May 21 by polgarp · Tagged in Resource and Course
Original link: https://medium.freecodecamp.org/every-single-machine-learning-course-on-the-internet-ranked-by-your-reviews-3c4a7b8026c0
Excellent list of courses related to machine learning. Great place to start if you wish to improve your AI skills.
Link to post
Posted on 2018 May 21 by polgarp · Tagged in Computing power and Statistics
Original link: https://blog.openai.com/ai-and-compute/
This is a nice statistics about the increase of computing power used in AI tech over the past years. Shows nicely one of the driving force behind the rise of AI tech in recent years.
Link to post
Posted on 2018 May 21 by polgarp · Tagged in AI limits
Original link: https://www.nytimes.com/2018/05/18/opinion/artificial-intelligence-challenges.html
Nice piece on the limitations of current AI tech (mostly machine learning). If we consider the goal to have human-level, so general AI in our hands.
Link to post
Posted on 2018 May 21 by polgarp · Tagged in Opinion
Original link: https://www.fastcompany.com/40570350/5-ways-to-ensure-ai-unlocks-its-full-potential-to-serve-humanity
An opinion piece on which direction AI developments should go. It’s worth to read it, even if you don’t agree with all the points made. If you happen to work on an AI product, it’s worth to consider some of these questions, like how you design the basic premise of your product, something that augments or something that replaces people.
Link to post
Posted on 2018 May 13 by polgarp · Tagged in Ethics
Original link: https://www.eff.org/deeplinks/2018/05/math-cant-solve-everything-questions-we-need-be-asking-deciding-algorithm-answer
Pretty cool guide from EFF on what circumstances could be a machine learning algorithm (for decision making) be considered. The additional explanations make them also quite actionable.
Link to post
Posted on 2018 May 13 by polgarp · Tagged in Google, Google Duplex, and Ethics
Original link: https://techcrunch.com/2018/05/10/duplex-shows-google-failing-at-ethical-and-creative-ai-design/
A longer article detailing the ethical concerns with Google Duplex, mostly coming to tricking humans with AI tech is not ethical.
Link to post
Posted on 2018 May 13 by polgarp · Tagged in Security and Privacy
Original link: https://www.nytimes.com/2018/05/10/technology/alexa-siri-hidden-command-audio-attacks.html
Fascinating article about attacks done via telling commands to smart systems that the humans cannot here. Especially the part about subliminal messages to machines not being illegal. A good reminder to think about privacy a lot when designing such systems.
Link to post
Posted on 2018 May 13 by polgarp · Tagged in Face recognition and AI fail
Original link: https://www.theverge.com/2018/5/6/17324496/south-wales-police-automated-facial-recognition-false-positives-privacy-security
Police used a failing facial recognition system that mostly had false positives and call it a success. So what about the false negatives? This is why you want to carefully balance the trade offs with your system.
Link to post
Posted on 2018 May 12 by polgarp · Tagged in AI design
Original link: https://petewarden.com/2018/05/07/why-ml-interfaces-will-be-more-like-pets-than-machines/
Some nice points in this article describing how an AI product would work, especially as an idea for interaction design. Probably this is just one model of interaction, but this would lead to a more personal connection to tech. And more anthropomorphism in how people perceive technology.
Link to post
Posted on 2018 May 12 by polgarp · Tagged in Google and Google Duplex
Original link: https://www.theverge.com/2018/5/9/17335710/google-duplex-phone-call-ai-assistant-service-industry
A slightly pessimistic view on how Duplex would change human-human interactions, that is ultimately resulting in less connection to services consumed. Good design is environmentally-friendly, as I see in this case, good design should also mind people affected.
Link to post
Posted on 2018 May 12 by polgarp · Tagged in Google
Original link: https://www.blog.google/topics/innovation-technology/io18-overview/
Sundar Pichai’s letter to this year’s I/O gives some interesting insights on the direction Google is going with their use of AI tech. One hand it’s just more technology will solve more problems, but all the product initiatives do seem to point into a shared vision.
Link to post
Posted on 2018 May 12 by polgarp · Tagged in Google and Google Duplex
Original link: https://www.theverge.com/2018/5/9/17334658/google-ai-phone-call-assistant-duplex-ethical-social-implications
Some predictions how Duplex might pan out, maybe even the comparison to Facebook M is right. The tech is just not really ready. Yet.
Link to post
Posted on 2018 May 12 by polgarp · Tagged in Google, Google Duplex, Research, and Assistant
Original link: https://ai.googleblog.com/2018/05/duplex-ai-system-for-natural-conversation.html
The announcement for Google Duplex, an extension of Assistant to make restaurant reservations and similar tasks over the phone. Even if it’s very limited, it does seem to work fine, and there a few nice details on the tech.
Link to post
Posted on 2018 May 12 by polgarp · Tagged in Facebook, Resource, and Development
Original link: https://facebook.ai/
A list of tools from Facebook to develop your AI solution. They have released some of their tools in the past, now it’s collected in one place.
Link to post
Posted on 2018 May 12 by polgarp · Tagged in Google, Google Clips, and Case study
Original link: https://ai.googleblog.com/2018/05/automatic-photography-with-google-clips.html
Cool case study on Google Clips. I liked especially how the data was labelled, instead of rating individual shots, comparison is much easier for even humans to perform and leads to better results.
Link to post
Posted on 2018 May 12 by polgarp · Tagged in Resource
Original link: https://aicheatsheet.comuzi.xyz/
A collection of AI tech terms and concepts, simply explained.
Link to post
Posted on 2018 May 12 by polgarp · Tagged in Google
Original link: https://www.blog.google/topics/developers/all-io18-announcements/
Google announced many interesting things this year on I/O. The list is quite telling, few product categories and AI as a separate group. Seeing as AI solutions appear everywhere, I wonder how long will be this tech a separate group and not just included in the other categories.
Link to post
Posted on 2018 May 7 by polgarp · Tagged in Google, Curation, and YouTube
Original link: https://gizmodo.com/youtube-basically-admits-algorithms-arent-such-good-bab-1825564100
Very interesting. So recommendation with machine learning resulted in some scary videos shown to children, a type error that shouldn’t be acceptable. Now that YouTube rolls out the feature to make human curation a choice, this shows making AI tech right is hard. Especially if you have users who have a different agenda than you.
Link to post
Posted on 2018 May 7 by polgarp · Tagged in Artificial general intelligence and Fun
Original link: https://kottke.org/18/04/the-lebowski-theorem-of-machine-superintelligence
Though more like fun (why would a general machine intelligence bother with killing humans, and not just leaning back), I like how this idea highlights that we assume the machine intelligence created by humans will be also similar to humans in how it thinks.
Link to post
Posted on 2018 May 7 by polgarp · Tagged in Business
Original link: https://techcrunch.com/2018/03/15/silicon-valley-companies-are-undermining-the-impact-of-artificial-intelligence/
While it’s true tech companies hoarding AI talent makes other industries adopting AI tech slower, it’s also true that there are huge opportunities to target more traditional industries with AI based solutions. Great opportunities for designers to find these problems and dream up solutions.
Link to post
Posted on 2018 May 7 by polgarp · Tagged in Facebook, Prediction, Ethics, and Dark pattern
Original link: https://theintercept.com/2018/04/13/facebook-advertising-data-artificial-intelligence-ai/
Overall it’s not that surprising Facebook can predict future behaviour, and only mildly surprising this didn’t result in an ethics stop for them, rather in a new source of revenue. At least this helps us exploring what would dark pattern mean in the case of AI tech.
Link to post
Posted on 2018 May 7 by polgarp · Tagged in Accessibility
Original link: http://www.fast.ai/2018/04/10/stanford-salon/
In design we mean accessibility as designing systems to be accessible to a wide range of users, regardless their backgrounds or capabilities. So it’s great there is effort made to make AI tech more accessible described in this article.
Link to post
Posted on 2018 May 6 by polgarp · Tagged in Product, Machine learning, and Tool
Original link: https://lobe.ai/
“Lobe is an easy-to-use visual tool that lets you build custom deep learning models, quickly train them, and ship them directly in your app without writing any code.” This seems to be the most accessible machine learning tool yet, might be a cool way for even designers to play around.
Link to post
Posted on 2018 May 6 by polgarp · Tagged in Machine learning
Original link: https://www.sciencemag.org/news/2018/05/ai-researchers-allege-machine-learning-alchemy
So machine learning (and probably all sophisticated AI tech) is complex, and very few people grasp it. So most people default to cargo-cult like usage. All the more reason designers strive to understand at least some of the principles, so they are able tell bullshit from the real thing.
Link to post
Posted on 2018 April 30 by polgarp · Tagged in Fake news
Original link: https://www.vox.com/science-and-health/2018/4/20/17109764/deepfake-ai-false-memory-psychology
Nice summary on the effects of fake videos, in particular how they may change a person’s memory of past events. Memories are not a fixed image people remember, this is well-known by design researchers as something that might bias research results. But now we see more and more examples where memories could be intentionally falsified, a completely new category of dark patterns.
Link to post
Posted on 2018 April 30 by polgarp · Tagged in Fashion, Amazon, Curation, and Algorithmic culture
Original link: https://www.racked.com/2018/4/17/17219166/fashion-style-algorithm-amazon-echo-look
“How good of a tastemaker can a machine ultimately be?” Loved this essay on algorithmic fashion and the broader questions of how AI will change what style the clothes should be designed in - and how algorithms change what fashion is. One power of AI tech is how it could fundamentally change big social and cultural constructs like fashion, with very few design choices that made sense at the time.
Link to post
Posted on 2018 April 30 by polgarp · Tagged in Resource and Dark pattern
Original link: https://github.com/daviddao/awful-ai
Nice list of AI applications that are in some way bad. Dark patterns (or more recently asshole patterns) will be soon referring features like this, and not only a choice of UI that keeps you from unsubscribing.
Link to post
Posted on 2018 April 29 by polgarp · Tagged in Report, Privacy, and Policy
Original link: https://privacyinternational.org/sites/default/files/2018-04/Privacy%20and%20Freedom%20of%20Expression%20%20In%20the%20Age%20of%20Artificial%20Intelligence.pdf
Nice report on the impact of AI tech on privacy.
Link to post
Posted on 2018 April 29 by polgarp · Tagged in History and Artificial general intelligence
Original link: http://rodneybrooks.com/forai-the-origins-of-artificial-intelligence/
Nice essay about the origins of artificial intelligence research, how the early researchers intended to target human level intelligence, and where it stands today. Spoilers - we are still far away.
Link to post
Posted on 2018 April 29 by polgarp · Tagged in AI chips
Original link: https://twitter.com/jwangARK/status/987432809127120896
Nice overview on all the companies currently working on AI chips. There are now at least 19 of them.
Link to post
Posted on 2018 April 23 by polgarp · Tagged in UK, Policy, Ethics, and Report
Original link: https://www.parliament.uk/business/committees/committees-a-z/lords-select/ai-committee/news-parliament-2017/ai-report-published/
A report on how AI should be developed in the UK by the Parliament. Some interesting recommendations, covering more or less what other institutions and governments already issued. Maybe the focus on ethics seems kind of specific. A little bit strange how more and more governments issue such recommendations, as technology impact on the broader society is difficult to predict.
Link to post
Posted on 2018 April 23 by polgarp · Tagged in AI design and Wizard of Oz
Original link: https://towardsdatascience.com/should-your-customers-be-conned-by-a-human-or-ai-6a87fbecdefe
Nice overview on using the Wizard of Oz method for prototyping AI products - and there is also the reflection how this compares to interfaces driven by humans (concierge methods). We still have too few methods to prototype AI experiences, so insights on the WoO methods are always interesting.
Link to post
Posted on 2018 April 23 by polgarp · Tagged in Report, Business, and Product
Original link: https://www.mckinsey.com/global-themes/artificial-intelligence/notes-from-the-ai-frontier-applications-and-value-of-deep-learning
Very comprehensive report on AI tech in the different industries and it’s potential, especially if you are looking to create AI products, or want to include AI tech into your company.
Link to post
Posted on 2018 April 23 by polgarp · Tagged in Research, Physics, and Chaos theory
Original link: https://www.quantamagazine.org/machine-learnings-amazing-ability-to-predict-chaos-20180418/
This is indeed stunning results. Lots of interesting systems are also chaotic, so predicting how they will behave will give us almost power over the future, for example when thinking about predicting weather or earthquakes.
Link to post
Posted on 2018 April 23 by polgarp · Tagged in Google and Human-centered AI
Original link: https://www.blog.google/perspectives/fei-fei-li/how-make-ai-good-for-people/
Love this sentence “No technology is more reflective of its creators than AI.” in this article about how to make AI more human, or human centered. I think approaching AI this way is the only sensible way - how to create technology in any other way if not as human centered?
Link to post
Posted on 2018 April 23 by polgarp · Tagged in Google, AI design, and Case study
Original link: https://design.google/library/designing-and-learning-teachable-machine/
Cool article from Google on how they created their Teachable Machine toy. We still don’t have many interaction patterns for AI applications, so such case studies are a good way to learn
Link to post
Posted on 2018 April 23 by polgarp · Tagged in AI terminology
Original link: https://medium.com/@mijordan3/artificial-intelligence-the-revolution-hasnt-happened-yet-5e1d5812e1e7
There are lots of confusion on the different topics inside AI tech, and even more notably on the term artificial intelligence itself - and this article clears these up mostly. A good read if you look to understand the whole spectrum of ideas under this umbrella term.
Link to post
Posted on 2018 April 23 by polgarp · Tagged in AI design
Original link: https://uxstudioteam.com/ux-blog/ai-ux/
Nice set of principles for designing AI products.
Link to post
Posted on 2018 April 23 by polgarp · Tagged in Bias, Ethics, and Job automation
Original link: https://medium.com/aclu/a-i-engineers-must-open-their-designs-to-democratic-control-b596496b4599
Nice summary on how current AI tech might impact the broader society around us, and some of the efforts looking to explore and mitigate the potential negative aspects. I liked the term “extended intelligence”, as many AI tools being developed indeed feel like more just enhancing human intelligence and not just being intelligent on it’s own. But I’m confused a little bit on the conclusion - there is already a discipline for creating technologies in a way that has a broader right impact, it’s called design.
Link to post
Posted on 2018 April 22 by polgarp · Tagged in Overview and Artificial general intelligence
Original link: http://mattturck.com/frontierai/
Nice overview article on current AI research efforts, mostly on the foundations and directions towards the general intelligence - though the initial questions doesn’t really gets answered. So in summary: far.
Link to post
Posted on 2018 April 22 by polgarp · Tagged in Institute and Research
Original link: https://ainowinstitute.org/
“The AI Now Institute at New York University is an interdisciplinary research center dedicated to understanding the social implications of artificial intelligence.” - seems interesting that more of these organizations pop up now.
Link to post
Posted on 2018 April 15 by polgarp · Tagged in Research and Surprise
Original link: http://aiweirdness.com/post/172894792687/when-algorithms-surprise-us
One of the most interesting aspect of working with AI tech is that it can exhibit unexpected, novel behaviour, like the examples in this article. In a way problems for such algorithms should be designed as you would design for a human. Only when machines do strange things, they do it differently then humans. So a challenge for designing products with AI tech is that now both parties, so humans and machines can do unexpected things.
Link to post
Posted on 2018 April 15 by polgarp · Tagged in Sci-fi and Analysis
Original link: https://scifiinterfaces.com/2018/04/09/untold-ai/
Excellent, an analysis on AI in sci-fi movies / TV shows. Or at least the first post in a series. Since sci-fi can be very inspirational on how future technology would feel like, would change things this is well worth reading if you are thinking about AI products.
Link to post
Posted on 2018 April 15 by polgarp · Tagged in Chine, Image recognition, Surveillance, and SenseTime
Original link: https://www.theverge.com/2018/4/11/17223504/ai-startup-sensetime-china-most-valuable-facial-recognition-surveillance
If you have wondered if AI would have an impact outside of tech, this is an excellent sign. The scale and kind of surveillance SenseTime can offer is something out of a sci-fi, but may be soon available to most governments.
Link to post
Posted on 2018 April 15 by polgarp · Tagged in Algorithmic warfare, Military AI, and Drone
Original link: https://www.theverge.com/2018/4/12/17229150/pentagon-project-maven-ai-google-war-military
Nice article detailing currently known Pentagon efforts on AI weapons, or more precisely autonomous drones built with AI tech. Just as any technology, AI tech can be designed to use for good or evil. Maybe if there is a ban on the military use of AI based weapons, it will also change how we think on using AI in commercial products.
Link to post
Posted on 2018 April 15 by polgarp · Tagged in Google, Experiment, and Search
Original link: https://books.google.com/talktobooks/
Interesting experiment from Google, allowing to ask natural language question giving back results from books. The results are quite good. Gets me thinking on how this could be expanded to a new experience interacting with books. For example besides reading a book, you could have a conversation with it, “what are you about?” and “what is the most important message?”.
Link to post
Posted on 2018 April 15 by polgarp · Tagged in Product, Art, and Co-creation
Original link: http://paintschainer.preferred.tech/index_en.html
“AI-Powered Automatic Colorization” for your line-art. Maybe not as good as a human colorist would do, but could serve well as tool to prototype colors and experiment.
Link to post
Posted on 2018 April 15 by polgarp · Tagged in Healthcare
Original link: https://www.theverge.com/2018/4/11/17224984/artificial-intelligence-idxdr-fda-eye-disease-diabetic-rethinopathy
This is significant, as it’s the first autonomous device that is approved, to be used without a specialist. The FDA is usually very rigorous on device approvals, one of the results being that less and less patient harm happens because of device errors and design errors.
Link to post
Posted on 2018 April 15 by polgarp · Tagged in Facebook
Original link: https://www.theverge.com/2018/4/13/17235042/facebook-mark-zuckerberg-ai-artificial-intelligence-excuse-congress-hearings
Nice summary on how Zuckerberg used the AI defence in his Congressional hearings. One of the most important things, AI in itself won’t solve anything. Well designed systems will solve issues. AI in itself most probably only magnify existing design issues through the scale and use of previous data.
Link to post
Posted on 2018 April 9 by polgarp · Tagged in Video and Resource
Original link: http://video.inspiring.ai/
Great list of videos related to AI + design, plenty of things to watch.
Link to post
Posted on 2018 April 9 by polgarp · Tagged in Art, Overview, Co-creation, and Augmented
Original link: https://qz.com/1023493/ai-will-be-the-art-movement-of-the-21st-century/
Nice overview on how AI tech could change art, and how a creative feedback loop can support artists in the future. AI tech is slowly becoming a material for creation as more and more projects emerge as shown here. But it’s still the wild west of interaction.
Link to post
Posted on 2018 April 8 by polgarp · Tagged in Product and Machine learning tool
Original link: https://clutch.ai/
“WorkBenchAI brings the power of predictive models to your domain with Zero Code interpretable ML.” It’s great there are more and more tools to play with ML, all the better for designers exploring this new “material”.
Link to post
Posted on 2018 April 8 by polgarp · Tagged in Google, Research, Ramen, and Fun
Original link: https://www.blog.google/topics/machine-learning/noodle-machine-learning-can-identify-ramen-shop/
Seems more fun than useful, a machine learning algorithm identifying which bowl of ramen came from which restaurant. Still interesting as far as how tiny details can be targeted by ML.
Link to post
Posted on 2018 April 8 by polgarp · Tagged in Bots and Lawyer
Original link: https://www.recode.net/2018/4/5/17119956/parking-ticket-property-tax-revenue
Nice overview on the rise of lawyer bots. I love these stories, as it shows how AI tech (or maybe even simpler bots with very little intelligence) can enable new product categories. In this case helping people without access to a lawyer (or money) to get help at scale.
Link to post
Posted on 2018 April 8 by polgarp · Tagged in Deepfakes, Fake news, and Research
Original link: https://www.fastcompany.com/40551971/can-new-forensic-tech-win-war-on-ai-generated-fake-images
AI generated fake news, and especially generated fake media, so pictures and videos start to have a real impact. This article shows a possible way out of it, but having the tech capability of detecting fake news is not the same as enabling people to realize when they see something they shouldn’t believe. Some very interesting design problems here.
Link to post
Posted on 2018 April 8 by polgarp · Tagged in France, AI strategy, and Report
Original link: https://www.aiforhumanity.fr/en/
The French AI strategy Macron was talking about earlier. Some nice details and proposals. It seems they aim to fix some of the more apparent issues like bias and drive the creation of good products.
Link to post
Posted on 2018 April 2 by polgarp · Tagged in Fake news and Overview
Original link: https://medium.com/@francois.chollet/what-worries-me-about-ai-ed9df072b704
Good overview on how AI tech enables manipulation on scale - and a few ideas on what to do about it. Mostly along the line that we should build some kind of anti-Facebook. Much seems to depend on the way AI tech is designed, for example transparent goals for the algorithms, opposed to the opaque one from current companies. And easy to understand controls to modify these.
Link to post
Posted on 2018 April 2 by polgarp · Tagged in Self-driving vehicles and Agentive design
Original link: https://www.ben-evans.com/benedictevans/2018/3/26/steps-to-autonomy
Some very interesting points on the deployment of self-driving cars by Benedict Evans. The autonomy level of a car means different things in different contexts, so there is probably multiple types of interaction styles such a car has. Driving a car might be more complicated in the future than using one of the bigger software pieces today - if it wasn’t designed well.
Link to post
Posted on 2018 April 2 by polgarp · Tagged in Art and Experiments
Original link: https://twitter.com/DrBeef_
Some fantastic art pieces generated by machine learning. Especially the ones showing people, something like Picasso and Dali combined.
Link to post
Posted on 2018 April 2 by polgarp · Tagged in Toy, Kids, and Learning
Original link: https://machinelearningforkids.co.uk/#!/about
Nice little project to use machine learning is Scratch projects, ideal for children to learn about this tech.
Link to post
Posted on 2018 April 2 by polgarp · Tagged in France and AI politics
Original link: https://www.wired.com/story/emmanuel-macron-talks-to-wired-about-frances-ai-strategy/
So France officially also joined the game, very few countries talk about AI on such a high level.
Link to post
Posted on 2018 April 2 by polgarp · Tagged in Fake news and Image manipulation
Original link: https://giorgiop.github.io/posts/2018/03/17/AI-and-digital-forgery/
Interesting article on the dangers of fake news, and two possible ways to fix it. Both seem to rely on people actually trying to work with them on their own. So it seems a huge challenge on how to design systems people could easily use, something that encrypted email wasn’t able to solve in the past 45 years.
Link to post
Posted on 2018 April 2 by polgarp · Tagged in China, Surveillance, and Overview
Original link: https://www.theatlantic.com/magazine/archive/2018/04/big-in-china-machines-that-scan-your-face/554075/
Nice overview on some of the efforts of the Chinese government to use machine learning to spy on it’s citizens.
Link to post
Posted on 2018 April 1 by polgarp · Tagged in Philosophy, Ethics, and Trolley problem
Original link: https://www.theatlantic.com/technology/archive/2018/03/got-99-problems-but-a-trolley-aint-one/556805/
A piece about the trolley problem, or more precisely on the different philosophical questions surrounding it. While currently this is specific to how to design self-driving cars, probably other product categories have similar design puzzles.
Link to post
Posted on 2018 March 26 by polgarp · Tagged in Research and Voice assistant
Original link: http://deliprao.com/archives/276
Female voices are harder to recognize than male due to technical reasons - and lack of adequate training data just reinforces this. Feels like a whole new class of accessibility issues will soon emerge.
Link to post
Posted on 2018 March 26 by polgarp · Tagged in Human-machine interaction
Original link: https://interestingengineering.com/flippy-the-burger-making-robot-was-fired-after-just-one-day-at-work
Interesting story, a robot introduced into a human workflow didn’t fit into, and now the restaurant thinks about retraining staff. Some deep interaction issues here, but the point is clear - even if you can introduce tech into a workflow done by humans you should work with the humans first.
Link to post
Posted on 2018 March 26 by polgarp · Tagged in Research, Product, and Parenthood
Original link: https://qz.com/1227955/muse-an-ai-powered-parenting-bot-wants-to-help-parents-help-their-children-succeed/
A bot that helps you to get better as a parent - by learning about your children’s abilities sounds interesting. And as the article mentions, touches depths of what makes us humans.
Link to post
Posted on 2018 March 26 by polgarp · Tagged in Job automation
Original link: https://medium.com/intuitionmachine/businesses-need-artificial-specialist-intelligence-not-the-other-kind-b0c285c1586e
Nice essay on the near future of job automation, especially about the points of automating street smarts most people have. When we design productivity software, we usually meant to give tools, so the users can focus on the creative tasks. What will happen if we start designing software that solves creativity too?
Link to post
Posted on 2018 March 26 by polgarp · Tagged in History
Original link: https://medium.com/@Saidov/12-breakthroughs-that-shaped-todays-artificial-intelligence-c38eb6dc4dd0
Nice overview on some of the ideas, technologies and products impacting today’s AI tech.
Link to post
Posted on 2018 March 25 by polgarp · Tagged in Healthcare and Case study
Original link: https://www.theverge.com/2018/3/21/17144260/healthcare-medicaid-algorithm-arkansas-cerebral-palsy
Many interesting details in this case study on a badly implemented algorithmic decision making system in Arkansas. Although not necessarily AI tech, it has very similar problems. Lot’s of issues that could have been solved with the right principles (like showing decision paths to users) and the right method (test the results of the algorithm).
Link to post
Posted on 2018 March 25 by polgarp · Tagged in Video games
Original link: https://www.theverge.com/2018/3/22/17150918/ea-dice-seed-battlefield-1-ai-shooter
EA training bots with ML might be not that new, as similar research was done in other games too. What’s more interesting, how these agents might be used in game development, for example game testing. Kind of like a testing tool. Maybe something like this can be done to speed up software development, for example augmenting usability testing.
Link to post
Posted on 2018 March 25 by polgarp · Tagged in Ethics
Original link: https://medium.com/intuitionmachine/the-dangers-of-artificial-intelligence-is-unavoidable-due-to-flaws-of-human-civilization-f9c131e65e5e
This article makes the point of AI is dangerous, since our society is not structured in a way to not make it dangerous. The is an interesting point, as designers think in exactly the opposite way: how to make technology fit into society and solve people’s problems. Designing an impactful technology (such as AI) is bound to change society in the end, so which comes first?
Link to post
Posted on 2018 March 19 by polgarp · Tagged in Machine learning and Development
Original link: https://petewarden.com/2018/03/19/the-machine-learning-reproducibility-crisis/
Interesting, though slightly technical description on why it’s hard to get the same result in machine learning by two different peoples. Reminds me a bit on the mostly messy way of how product design is done. This issues may also impact user experience down the line, so it’s at least useful to understand it.
Link to post
Posted on 2018 March 19 by polgarp · Tagged in Toolkit, AI design, and Workshop
Original link: http://iadesignkit.com/
Very cool toolkit to run your own workshop to define products / services with AI tech.
Link to post
Posted on 2018 March 19 by polgarp · Tagged in Philosophy and Design
Original link: https://medium.com/intuitionmachine/how-automation-and-artificial-intelligence-strips-us-of-our-humanity-352a9b92b873
Some interesting points raised, especially for designers. If we think meaningful and impactful work is needed to lead a full life, how does AI change this? Will it make us less or more happy. Tedious work should be eliminated, but maybe not everything - we already know this from UX. Challenging design question.
Link to post
Posted on 2018 March 19 by polgarp · Tagged in History and Automation
Original link: http://publicdomainreview.org/2016/05/04/frolicsome-engines-the-long-prehistory-of-artificial-intelligence/
Nice article on the history of automations, machines that some ways mimicked people or other real objects.
Link to post
Posted on 2018 March 18 by polgarp · Tagged in Comics, Fun, and xkcd
Original link: https://xkcd.com/1968/
Things to worry about with AI by Randall Munroe.
Link to post
Posted on 2018 March 18 by polgarp · Tagged in Movie
Original link: https://morehumanthanhuman.ai/
Sounds an interesting movie. “More Human Than Human explores what it means to live in the age of intelligent machines. In this personal, playful and at times dramatic quest, the filmmaker finds out how much of his creativity and human values are at stake as he builds his own robot to replace himself as a filmmaker.”
Link to post
Posted on 2018 March 18 by polgarp · Tagged in Ethics
Original link: https://medium.com/@mrtz/how-big-data-is-unfair-9aa544d739de
Great article on ethics, as it doesn’t just make a point on what should be done, but actually goes into detail how fairness is determined by the algorithms in machine learning. Some important points to keep in mind when striving to design an ethical AI product.
Link to post
Posted on 2018 March 18 by polgarp · Tagged in Google, Research, Co-creation, and Creativity
Original link: https://9to5google.com/2018/03/13/googles-nsynth-super-is-a-touchscreen-synthesizer-w-ai-to-create-new-sounds-video/
Co-creation with AI tech is an interesting topic, as a lot of opportunities seem possible for artists, but there few working examples yet. Though this synthesizer from Google is not for sale, you can build one yourself based on the plans and play with it.
Link to post
Posted on 2018 March 18 by polgarp · Tagged in Talk and Product management
Original link: https://www.mindtheproduct.com/2018/03/diogo-coutinho-ai-real-world/
Some nice tips on AI in production in this talk by Diogo Coutinho at Mind the Product.
Link to post
Posted on 2018 March 18 by polgarp · Tagged in Report
Original link: https://designintech.report/
John Maeda’s annual report on design in the tech sector is anyways interesting for people designing products. But this year’s report also contains a lot of info on AI tech + design. Recommended.
Link to post
Posted on 2018 March 18 by polgarp · Tagged in Research, Experiment, and Conversation
Original link: http://collectivedebate.mit.edu/
Very interesting experiment, where you debate with an AI powered agent that provides counterpoint to your views. AI tech in the wild on social networks is often used to create filter bubbles, so it’s interesting to see a counterpoint to that. Besides the content the interaction is also well designed, kind a like in a video game.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Google, YouTube, and Machine learning
Original link: https://www.nytimes.com/2018/03/10/opinion/sunday/youtube-politics-radical.html
YouTube serving you ever more extreme content is not that surprising. This is kindof how such a machine learning supposed to work, guiding people to engagement directions that worked pretty well earlier.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Microsoft, Windows, and AI platform
Original link: https://www.theverge.com/2018/3/7/17089860/microsoft-windows-ml-windows-10-ai-platform
Windows lags a little bit behind MacOS in integrating Ml services. I wonder if this will drive the design of Windows apps.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Co-creation, Augmented creativity, and Procedural generation
Original link: https://www.theverge.com/2018/3/11/17099578/ai-generated-podcast-procedural-storytelling-art-sheldon-county
Procedurally generating stories is not that new, but both the complexity and the format seems interesting here. Maybe in the future personalised and uniquely generated entertainment will form a distinct category, would be interesting to see how this changes individual’s storytelling.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Research and Explainability
Original link: https://distill.pub/2018/building-blocks/
Explaining machine learning decisions is an important topic (especially if something goes wrong). This article proposes some ideas on this - and there are some really well done interactive examples to show how well they work.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Google, Research, and Art
Original link: https://www.blog.google/topics/arts-culture/machine-learning-meets-culture/
Three cool projects from Google, that lets you explore art in a different way.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Collection, AI tools, and Co-creation
Original link: https://www.producthunt.com/@jgage718/collections/machine-learning-for-creatives
Awesome tool list - “Here are 15 products to help you design and build beautiful things, powered by Machine Learning”.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Human centred AI and AI design
Original link: https://www.nytimes.com/2018/03/07/opinion/artificial-intelligence-human.html
Interesting opinion piece on what’s essentially the design (maybe even UX design) of AI tech. I liked the idea of how AI values are in essence the values of it’s human creators - this grabs the core of how designer bias is propagated in AI products.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Google, Collaboration, and Product
Original link: https://gsuiteupdates.googleblog.com/2018/03/find-shared-content-with-new-file.html
It’s always interesting to see, when Google rolls out a new feature with AI. Finding people who you want to collaborate at this moment seems useful, like an assistant helping you in your job. This seems like a minor improvement, that could have been also done with something simple, but apparently for Google AI tech is just as easy.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Augmented art, Art, Co-creation, and Overview
Original link: https://qz.com/1023493/ai-will-be-the-art-movement-of-the-21st-century/
Many nice ideas and some nice examples on how AI may become an other artistic tool. This is also happening in design, co-creation seems more exciting than plain automation.
Link to post
Posted on 2018 March 11 by polgarp · Tagged in Security and Adversarial
Original link: https://www.wired.com/story/ai-has-a-hallucination-problem-thats-proving-tough-to-fix/
If there is an incentive to game a system, there will be people who will game the system - even if the rewards are just knowing they could do it. Since AI tends to be deployed to be automatically used, there is bound to be examples where somethings is happening without the knowledge of an AI tech’s human owners.
Link to post
Posted on 2018 March 5 by polgarp · Tagged in Machine learning and Story
Original link: https://blog.evjang.com/2018/02/teacup-story.html
Teaching algorithms does seem very similar to how you would teach a small child. Minus children seem to have an uncanny ability to generalize from seeing objects once.
Link to post
Posted on 2018 March 5 by polgarp · Tagged in Bias
Original link: https://www.technologyreview.com/s/610275/meet-the-woman-who-searches-out-search-engines-bias-against-women-and-minorities/
It seems while creators of biased software could do something, and it’s probably their responsibility to make software the right way, it’s not their fault by default. The scale at which such algorithms, and AI tech in particular operates just magnify the society’s biases.
Link to post
Posted on 2018 March 4 by polgarp · Tagged in Experiment, Writing, and Creativity
Original link: https://www.robinsloan.com/voyages-in-sentence-space/
Cool idea, making “sentence gradients”. AI tech enables new forms of co-creation and creativity such as this one. I wonder what other concepts could be formed as gradients.
Link to post
Posted on 2018 March 4 by polgarp · Tagged in Bias and Overview
Original link: https://www.fastcompany.com/40536485/now-is-the-time-to-act-to-stop-bias-in-ai
Bias is a fact of life when working with AI tech, and it seems pretty hard to get rid of - even if you are committed to have a mostly bias free algorithm. Bias seems very similar to the problems design has (dark patterns included), for example while we may know diversity in the design team matters a lot, we rarely have truly diverse team. These are fine lines real product development threads.
Link to post
Posted on 2018 March 4 by polgarp · Tagged in Google and Course
Original link: https://www.blog.google/topics/machine-learning/learn-google-ai-making-ml-education-available-everyone/
Tons of learning materials from Google for beginners and advanced people. Pretty good resource which ever level you are.
Link to post
Posted on 2018 March 4 by polgarp · Tagged in Overview and Design tools
Original link: https://blog.marvelapp.com/ai-started-impact-work-designers/
A few nice tools and use cases how AI may or will help designers in the near future.
Link to post
Posted on 2018 March 4 by polgarp · Tagged in Research and AI control
Original link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2954173
This paper explores a simple question. Startups long disrupted business by moving fast. So what if we give the control of a business to an AI - that can scale up to move way faster than any human-owned business could possibly move?
Link to post
Posted on 2018 February 26 by polgarp · Tagged in Experiment and Deep learning
Original link: https://towardsdatascience.com/how-to-create-data-products-that-are-magical-using-sequence-to-sequence-models-703f86a231f8
Cool article, showing how easy it is to build simple automations on top of ML algorithms. Something like this would be useful in every design researcher’s toolbox when shifting through a large amount of qualitative data.
Link to post
Posted on 2018 February 26 by polgarp · Tagged in Overview
Original link: https://thinkgrowth.org/common-misconceptions-executives-have-about-ai-a4bee8f0de87
Would be fun to see which executives actually have these misconceptions. I wonder how many times somebody utters on brainstorming meetings “let’s use AI to solve this” though. Probably we will see a few years of “AI strategy” as we still see “mobile strategy”.
Link to post
Posted on 2018 February 25 by polgarp · Tagged in Resource and Ethics
Original link: https://medium.com/@eirinimalliaraki/toward-ethical-transparent-and-fair-ai-ml-a-critical-reading-list-d950e70a70ea
Cool link list on all things ethics.
Link to post
Posted on 2018 February 25 by polgarp · Tagged in AI apocalypse and Security
Original link: https://maliciousaireport.com/
Very through report on how AI can be misused by evil people. In case your long term worries included AI killing of humans, read this first.
Link to post
Posted on 2018 February 25 by polgarp · Tagged in Research, Healthcare, and Google
Original link: https://blog.verily.com/2018/02/eyes-window-into-heart-health.html
Remarkable research, predicting cardiovascular disease just based on looking at the eyes. The most interesting part is the attention map, that would show a doctor which parts of the image actually influenced the algorithm. This feels like also the right design for such an application: support the user with details of the recommendation the algorithm made.
Link to post
Posted on 2018 February 25 by polgarp · Tagged in AI design, Interview, and Capital One
Original link: https://www.fastcodesign.com/90160881/capital-one-vp-the-impact-ai-is-having-on-design-is-bananas
Interesting interview, especially that there are not that many companies where there is such a strong connection between AI and design.
Link to post
Posted on 2018 February 25 by polgarp · Tagged in Ethics, Government, and Policy
Original link: https://medium.com/@AINowInstitute/algorithmic-impact-assessments-toward-accountable-automation-in-public-agencies-bd9856e6fdde
There is lots of talk of ethics and designing algorithms in an ethical way, but there are few examples of something more substantial in terms of what we should do. I like this proposal as it’s more concrete and actionable than most recommendations or frameworks.
Link to post
Posted on 2018 February 19 by polgarp · Tagged in AI capabilities, Sustainability, and Tweet storm
Original link: https://twitter.com/TylerGlaiel/status/953845127281197058
Interesting tweet storm on the capabilities of AI tech. More specifically on why going for local optimum is hurting long term. One of the property of good design is long-lasting, that it provides value to the users also over long term. Working with this principle on AI tech might prove to be a challenge.
Link to post
Posted on 2018 February 19 by polgarp · Tagged in Statistics, Investment, and Analysis
Original link: http://files.pitchbook.com/website/files/pdf/PitchBook_1Q_2018_Analyst_Note_Real_Potential_for_AI.pdf
Nice analysis on the state of investments the AI market. I’m missing some numbers for other technologies and fields to compare against, but the overall impression is that it’s worth to pitch ideas and companies based on AI tech.
Link to post
Posted on 2018 February 19 by polgarp · Tagged in Trust and AI design
Original link: https://theconversation.com/people-dont-trust-ai-heres-how-we-can-change-that-87129
Trust is an interesting characteristic to design for, even with non-AI products where the behaviour of the computer is more or less predictable upfront and can be taken into account in design time. For AI tech based systems trust is more difficult as the final experience might be less predictable.
Link to post
Posted on 2018 February 19 by polgarp · Tagged in AI apocalypse and Psychology
Original link: http://quillette.com/2017/12/14/irrational-ai-nxiety/
Nice article detailing what are some of the psychological reasons behind fearing the AI apocalypse.
Link to post
Posted on 2018 February 19 by polgarp · Tagged in Predictive analytics, Government, and Use case
Original link: https://www.nytimes.com/2018/01/02/magazine/can-an-algorithm-tell-when-kids-are-in-danger.html
Products using this type of tech seems hard get right. Not because of the tech being inherently difficult, but rather the context and the many parties concerned. I liked this use case, since it showed how wide such a product’s introduction should go. Also makes you feel participatory design this easy: finding the right place to use the product and talking to everybody.
Link to post
Posted on 2018 February 18 by polgarp · Tagged in Neuroscience, Neurons, Research, Deep learning, and Neural networks
Original link: https://medium.com/intuitionmachine/neurons-are-more-complex-than-what-we-have-imagined-b3dd00a1dcd3
Some interesting insights on where the research on neural networks could go in the future.
Link to post
Posted on 2018 February 18 by polgarp · Tagged in Agriculture and China
Original link: https://www.theverge.com/2018/2/16/17019446/farming-ai-pig-tracking-china-alibaba
Face recognition for pigs.
Link to post
Posted on 2018 February 12 by polgarp · Tagged in Self-driving, Promotion, and Nissan
Original link: https://newsroom.nissan-global.com/releases/release-0a3c17000171b2fc662bfd35f7000704-180125-01-e
Maybe just a marketing gimmick, but in case you ever wondered, how certain AI tech will change how we interact with the world.
Link to post
Posted on 2018 February 11 by polgarp · Tagged in Fun
Original link: http://botnik.org/content/tng.html
Actually it’s not that bad, considering some episodes it can be almost real. Certainly better than some fanfics.
Link to post
Posted on 2018 February 11 by polgarp · Tagged in Fun
Original link: https://boingboing.net/2018/02/08/mad-magazine-exclusive-the.html
Posted on 2018 February 11 by polgarp · Tagged in AI apocalypse and Video gaming
Original link: https://www.techpowerup.com/240655/leaked-ai-powered-game-revenue-model-paper-foretells-a-dystopian-nightmare
While terminators killing people is a popular way of showing how the AI apocalypse would go off, this article shows a much more likely way of doing it: evil companies using AI tech to make people more addicted to technology. Ofcourse this is just a sales presentation for now, and product challenges are still a thing, but still possible.
Link to post
Posted on 2018 February 11 by polgarp · Tagged in Frontend development
Original link: https://medium.freecodecamp.org/how-you-can-train-an-ai-to-convert-your-design-mockups-into-html-and-css-cc7afd82fed4
“Should designers code?” “No, just be able to train neural networks.”
Link to post
Posted on 2018 February 11 by polgarp · Tagged in Centaur and Intelligence augmentation
Original link: https://jods.mitpress.mit.edu/pub/issue3-case
(One of) The most interesting questions these days in design is how humans and machine with AI tech should work together. While some or even most of the interaction may be similar, the underlying ideas might offer new or at least different perspectives. As this articles puts it, artificial intelligence augmentation might be a way to view this question.
Link to post
Posted on 2018 February 11 by polgarp · Tagged in Bias and Face recognition
Original link: https://www.nytimes.com/2018/02/09/technology/facial-recognition-race-artificial-intelligence.html
Not as surprising, bias in data will result bias in the algorithm. So let’s assume you are creating a product with face recognition - now you have to decide what errors will be acceptable for you. Much of product design is about trade offs, but getting good face recognition, even if errors only affect a small number of your users may be comparable to providing good accessibility.
Link to post
Posted on 2018 February 7 by polgarp · Tagged in Ethics and Talk
Original link: https://vimeo.com/209303271
“Design is applied ethics.” This statement exemplifies very well, why I think it’s important for designers working with AI tech to think about ethical questions. This talk by Cennydd Bowles offers a nice overview on ethical questions, and also provides some practical points, like ethical tests one can do on their design work.
Link to post
Posted on 2018 February 5 by polgarp · Tagged in Job automation and AI product development
Original link: http://marginalrevolution.com/marginalrevolution/2018/02/will-truckers-automated-comments.html
A thoughtful and interesting comment on Could Self-Driving Trucks Be Good for Truckers?. So some analysts make the mistake of thinking creating product is just about feasibility, when the broad user context should be also taken in account.
Link to post
Posted on 2018 February 5 by polgarp · Tagged in Ethics, Book, and Algorithmic cruelty
Original link: https://www.amazon.com/dp/1250074312/ref=cm_sw_r_tw_dp_U_x_CVmEAbMWVG6VR
Interesting book on algorithmic cruelty.
Link to post
Posted on 2018 February 5 by polgarp · Tagged in Research, Privacy, Data gathering, and Anonymous data
Original link: http://randomwalker.info/publications/precautionary.pdf
“Once released to the public, data cannot be taken back.” Anonymized personal data might be used in various ways, like you’d want to analyse your user’s data to create products based on them. But even if personally identifiable data is removed (like name or email address), what remains could be also used to identify people, unless the right methods used.
Link to post
Posted on 2018 February 4 by polgarp · Tagged in Overview, Psychiatry, and Healthcare
Original link: https://medium.com/neodotlife/computational-psychiatry-c05a32f20705
Interesting overview on current AI tech (promises) on the field of psychiatry. Very insightful, how general AI / machine learning might or might now work in a specific field, based on the underlying assumptions of the field. Like in this case: disorders are not that easy to define, and there might not be a systematic, generalizable pattern behind them, for an algorithm to find.
Link to post
Posted on 2018 February 4 by polgarp · Tagged in AI designer tool and Tool
Original link: https://www.fastcodesign.com/90157777/this-is-the-worlds-first-graphical-ai-interface
The article’s title might be over the top, but is actually interesting. Creating editors for a new type of interaction (or really for anything that should help people get creative) is always challenging. Feels like AI design tools will evolve similarly as HTML design tools did 15-20 years ago.
Link to post
Posted on 2018 February 4 by polgarp · Tagged in Product and Time tracking
Original link: https://timelyapp.com/memory-ai
Time tracking does seem to be one of the more tedious administrative task that is daily, needs to be done and requires precise completion to be useful. Memory AI promises to solve this neatly. I wonder how this additional data would affect one works. Reflection is a useful tool to aggregate the past, while forgetting details might help in seeing the bigger picture, so more data on your days doesn’t necessary mean also more knowledge. Or knowing what to do.
Link to post
Posted on 2018 February 4 by polgarp · Tagged in Hospitality and Overview
Original link: https://chatbotsmagazine.com/how-hotels-are-using-artificial-intelligence-to-provide-an-awesome-user-experience-466a70453b53
Nice, but brief overview how the hospitality vertical might use AI tech to enhance their service. I like how the article refers to improved service as better UX (or should it be ‘Guest eXperience’?). Plenty of opportunities for a vertical-focused startup.
Link to post
Posted on 2018 February 4 by polgarp · Tagged in Google and Travel
Original link: https://www.blog.google/topics/travel/get-help-cheaper-flights-and-potential-flight-delays-winter/
If flight delays can be predicted, there may be some systematic reason behind them. Nice from Google, to pressure airlines to fix these issues.
Link to post
Posted on 2018 February 4 by polgarp · Tagged in Chatbots and History
Original link: https://www.filfre.net/2011/06/eliza-part-1/
Eliza was the grandmother of chatbots, and reading about it’s history might put current developments in AI tech into a different perspective. It’s fascinating how even some very basic scripts could enable people to perceive their conversation with Eliza to be meaningful. Something to think about if you create chatbots.
Link to post
Posted on 2018 January 29 by polgarp · Tagged in Machine learning, Overview, and GDPR
Original link: https://medium.com/@giacomo_59737/the-delusions-of-neural-networks-f7085d47edb6
Very nice and insightful description of neural networks and machine learning to laymen. Also gives a good explanation why EU’s GDPR will make life of AI companies harder - you need to be able to explain what your algorithms are doing.
Link to post
Posted on 2018 January 29 by polgarp · Tagged in Overview and Business
Original link: https://techcrunch.com/2018/01/01/building-ai-systems-that-work-is-still-hard/
Nice overview on the possible landscape for AI companies. Tl;dr original research mostly happens at big companies as the tech is too complicated, proprietary teach gets open sourced fast, and much of the work will be along the last mile to the customers - just as with non AI companies. Plenty of opportunities for designers.
Link to post
Posted on 2018 January 29 by polgarp · Tagged in AI fail and Fun
Original link: https://syncedreview.com/2017/12/23/2017-in-review-10-ai-failures/
Some of the examples are quite fun. And all of them shows how even big name products will have issues - AI tech is hard.
Link to post
Posted on 2018 January 28 by polgarp · Tagged in Google, AI UX, and Human-centered machine learning
Original link: https://design.google/library/ux-ai/
Extremely interesting article by Google, explaining their ML design principles through a case study of their new Clips camera. As machine learning as a design material is still so new, I think most teams are still experimenting what works and what doesn’t, especially process wise, but they at least have already working products out there.
Link to post
Posted on 2018 January 28 by polgarp · Tagged in Research, DeepMind, and Psychology
Original link: https://deepmind.com/blog/open-sourcing-psychlab/
“we developed Psychlab, a platform built on top of DeepMind Lab, which allows us to directly apply methods from fields like cognitive psychology to study behaviours of artificial agents in a controlled environment. Today, we are also open-sourcing this platform for others to use.” Very cool, especially since most concepts from the field of design are essentially applied cognitive psychology.
Link to post
Posted on 2018 January 28 by polgarp · Tagged in Privacy
Original link: https://www.theverge.com/2018/1/23/16907238/artificial-intelligence-surveillance-cameras-security
Some of the examples sound like a sci-fi (“Computer, show me all passersby…”) or promises to be one soon, and it’s though to decide if we are heading towards an utopia or dystopia. But in either case as the article points out, there are still many challenges in actually deploying these systems in way that they get at least most requests right.
Link to post
Posted on 2018 January 21 by polgarp · Tagged in Job automation and Overview
Original link: https://www.nytimes.com/2017/12/11/style/robots-jobs-children.html
Interesting take on the job automation take - what should your children study right now, so their job won’t be automated. Though, since we may expect a huge ratio of jobs to be automated even before our children start school.
Link to post
Posted on 2018 January 21 by polgarp · Tagged in Fun and Typography
Original link: https://medium.com/swlh/why-sans-serif-is-a-problem-for-ai-efeebae8df5
“Yes, this just became about typography.”
Link to post
Posted on 2018 January 21 by polgarp · Tagged in Ethics
Original link: https://ia.net/topics/who-serves-whom/
I consider ethics an important question when thinking about AI tech, since design in it’s core humanistic, and those implicit values need to be made explicit when thinking about the power and scale of AI tech. This article explores quite a lot of points in this topic, especially concerning human control on machine and it’s limits.
Link to post
Posted on 2018 January 21 by polgarp · Tagged in Google, Machine learning, and Introduction
Original link: https://docs.google.com/presentation/d/1kSuQyW5DTnkVaZEjGYCkfOxvzCqGEFzWBy4e9Uedd9k/preview?imm_mid=0f9b7e&cmp=em-data-na-na-newsltr_20171213#slide=id.g168a3288f7_0_58
Nice introduction for machine learning - contains many nice examples and also some pointers to get started.
Link to post
Posted on 2018 January 21 by polgarp · Tagged in Research and Machine learning
Original link: https://www.wired.com/story/crime-predicting-algorithms-may-not-outperform-untrained-humans/
Interesting results. One way of coming up with ideas for new machine learning products is to think about problems where human intuition and experience solves the problem. Ofcourse the resulting product won’t be better then a human would be, only better at scale. What’s strange here is that the algorithm would be only as good as untrained humans. Issues + bias in data and no validation on results maybe?
Link to post
Posted on 2018 January 21 by polgarp · Tagged in Google and Infrastructure
Original link: https://www.blog.google/topics/google-cloud/cloud-automl-making-ai-accessible-every-business/
Pretty exciting news. With this creating machine learning products, or just building prototypes will be much easier, maybe even for designers.
Link to post
Posted on 2018 January 21 by polgarp · Tagged in Amazon
Original link: https://www.fastcompany.com/40518124/amazon-go-store-opening
As far as product / services go, this is an astonishing application of AI tech, even with the details obscured. In the end anyway the experience is what counts. Shows very clearly the power of AI tech - we can design for a bigger class of problems.
Link to post
Posted on 2018 January 21 by polgarp · Tagged in Netflix, Personalization, and Content personalization
Original link: https://medium.com/netflix-techblog/artwork-personalization-c589f074ad76
Pretty cool description on how Netflix personalizes movie artwork to match you - and drive you to watch ever more movies. Besides the interesting details this gives a nice perspective on the depth of content personalisation you can think about when creating a content heavy product.
Link to post
Posted on 2018 January 15 by polgarp · Tagged in UI design, Code generation, and Uizard
Original link: https://hackernoon.com/teaching-machines-to-understand-user-interfaces-5a0cdeb4d579
Pretty cool, especially if I think of the tediousness of creating and iterating on trivial interfaces. We would only need some simple mechanism to link together screens and we would have a whiteboard to hi-fi prototype tool. This may answer the age old question, “should designers code?”.
Link to post
Posted on 2018 January 15 by polgarp · Tagged in Privacy and Data entry
Original link: https://www.wired.com/story/not-always-ai-that-sifts-through-sensitive-info-crowdsourced-labor/
Fake it till you make it gets a whole new meaning, if you realize, your private data is accessed by random people with very limited willingness to keep your privacy. Something to think about if you design data collection for your product.
Link to post
Posted on 2018 January 15 by polgarp · Tagged in Agency and Overview
Original link: https://medium.com/@jackkrupansky/intelligent-entities-principals-agents-and-assistants-8353639a4092
If you design systems with AI tech, you might start to think about what’s the difference between the different actors, what makes a solution intelligent and how the human and computer elements fit together. This extremely long article gives a very detailed overview on some of these concepts.
Link to post
Posted on 2018 January 15 by polgarp · Tagged in Research, Agriculture, and Chickens
Original link: https://www.scientificamerican.com/article/fowl-language-ai-decodes-the-nuances-of-chicken-ldquo-speech-rdquo/
Fascinating research project, understanding chickens better with AI tech to make them happier. Just another example where human intuition can be automatized - and also how hard it is in production.
Link to post
Posted on 2018 January 14 by polgarp · Tagged in Google, Google Photos, Gorilla, and Computer vision
Original link: https://www.wired.com/story/when-it-comes-to-gorillas-google-photos-remains-blind/
So instead of fixing the whole recognizing-people-as-gorillas issue Google decided to just remove the tag gorilla. If anything this just shows that while machine learning has great promise actually using this technology in a production environment may create unexpected and really hard issues that takes tons of effort to solve. And tons of time.
Link to post
Posted on 2018 January 14 by polgarp · Tagged in Interview
Original link: https://machinelearnings.co/machine-learning-wont-reach-its-potential-without-the-human-element-5b030051cd9d
Interesting interview with Sarah Catanzaro of Amplify Partners, a VC focused mostly on machine learning. I liked the idea best how machine learning sheds light on our existing biases, which in turn could be used to design better services for people.
Link to post
Posted on 2018 January 14 by polgarp · Tagged in Ethics, Book, and Ethnography
Original link: https://points.datasociety.net/beyond-the-rhetoric-of-algorithmic-solutionism-8e0f9cdada53
Machine learning, or really an AI tech will change society at large just as big technological innovations did in the past (I’m thinking on web, mobile, social networks), how it happens is up to us, people designing these products. This seems and excellent book coming from an ethnography angle showing where the changes are heading.
Link to post
Posted on 2018 January 14 by polgarp · Tagged in AI attack and Adversarial
Original link: https://thenextweb.com/artificial-intelligence/2018/01/11/ai-learns-how-to-fool-text-to-speech-thats-bad-news-for-voice-assistants/
Sounds like a pretty neat technique, being able to make a voice assistant to do anything by disguising voice commands from a human’s ears. While it could be used as a kind of flash bang for computers, other ideas also come to mind, like suggesting a device to download an app, open a page or simply adding a song to a playlist.
Link to post
Posted on 2018 January 14 by polgarp · Tagged in AI desgin and Tips
Original link: https://blog.prototypr.io/a-designers-guide-to-the-future-5b68566e36bf
Some nice tips on getting started with designing AI tech. One great advice is to actually try to use such services, as the “material” for AI tech is still so new, we have to get a feeling for it.
Link to post
Posted on 2018 January 11 by polgarp · Tagged in Comics and Fun
Original link: https://www.smbc-comics.com/comic/evil-ai
“Wanna conquer and enslave humanity?”
Link to post
Posted on 2018 January 11 by polgarp · Tagged in Ethics, Principles, and Overview
Original link: http://robohub.org/a-round-up-of-robotics-and-ai-ethics-part-1-principles/
When thinking about ethics and about ethics for AI people frequently summarize their thought with list of priorities this article lists all known principle collection starting with Asimov’s laws for robotics. Nice to read through and think about the various perspectives.
Link to post
Posted on 2018 January 8 by polgarp · Tagged in Adversary system, Computer vision, Google, and Research
Original link: https://techcrunch.com/2018/01/02/these-psychedelic-stickers-blow-ai-minds/
This might be the first practical application of attacking a computer vision system. I assume more and more of similar tricks will emerge against computer vision systems registering you in the real world, just as adblockers emerged in response to the ad tracking on websites. Plenty of interesting problems to design for and against, you might want to create systems that has explicit benefits to people without ripping of their data to not face users having this against you.
Link to post
Posted on 2018 January 8 by polgarp · Tagged in Fun, Machine learning, and Experiments
Original link: http://aiweirdness.com/post/169049462997/ten-new-applications-for-neural-networks
A collection of fun machine learning experiments by Janelle Shane showing how machine learning can be used for creative endeavours.
Link to post
Posted on 2018 January 7 by polgarp · Tagged in AI chips and Overview
Original link: https://fieldguide.gizmodo.com/what-do-the-ai-chips-in-new-smartphones-actually-do-1820913665
Nice overview on what the AI chips in current smartphones are actually doing. In short, they just make basic machine learning tasks faster
Link to post
Posted on 2018 January 7 by polgarp · Tagged in General AI
Original link: https://medium.com/@francois.chollet/the-impossibility-of-intelligence-explosion-5be4a9eda6ec
As far as arguments against “superintelligent” AI this is a very thorough article, mostly comparing possibilities to the human angle. Still barely touches issues more practical in nature like how software bugs might work in such a system.
Link to post
Posted on 2018 January 7 by polgarp · Tagged in Introduction, Video, and Machine learning
Original link: https://www.youtube.com/watch?v=R9OHn5ZF4Uo
Cute introduction video to machine learning by CGP Grey. While the explanation uses very basic terms, it goes into quite the depths. Bonus points for representing collecting data as a design researcher observing a user.
Link to post
Posted on 2018 January 7 by polgarp · Tagged in Google, Research, Voice interface, and Guideline
Original link: https://research.googleblog.com/2017/12/evaluation-of-speech-for-google.html
Nice evaluation guidelines for voice interfaces for Google Assistant. It’s clear that for machine learning results we also need some qualitative evaluation, to make sure it meets human needs beyond numbers.
Link to post
Posted on 2018 January 7 by polgarp · Tagged in Research, Video games, and AI apocalypse
Original link: https://www.theverge.com/2018/1/5/16853628/civilization-v-ai-risk-mod-cser
Studying research questions through video games is always fascinating, as it also opens up to a wider audience and can result in unexpected results. Even if this is just a limited context, playing with the core ideas might be interesting.
Link to post
Posted on 2018 January 7 by polgarp · Tagged in AI policy and AI ethics
Original link: http://rodneybrooks.com/aiml-is-not-uniquely-powerful-enough-to-need-controlling/
Nice counterargument for policy making on AI research. There seems to be no agreement even between AI researchers on what is possible at all, what should be possible and what needs regulation if anything at all. Also policy making for AI at this point might be YAGNI, as turning theories and ideas onto products at scale is much harder than most people realize.
Link to post
Posted on 2018 January 5 by polgarp · Tagged in AI apocalypse and Singularity
Original link: https://freedom-to-tinker.com/2018/01/03/why-the-singularity-is-not-a-singularity/
Many of the arguments about how we are heading towards an AI apocalypse is based on the concept of singularity. This series of articles dismantles this concept, and is overall an interesting read.
Link to post
Posted on 2018 January 5 by polgarp · Tagged in Content generation
Original link: https://www.wired.com/story/future-of-artificial-intelligence-2018/
The most interesting task seems to be when you create the tools that enable users to generate potentially fake content. Photoshop didn’t allow editing banknotes for some time to make counterfeiting harder, I wonder if such safeguards make sense this time. Or if they make sense, as machine learning seems so cheap.
Link to post
Posted on 2017 December 18 by polgarp · Tagged in Ethics, AI design, and Overview
Original link: https://hackernoon.com/the-real-danger-of-artificial-intelligence-its-not-what-you-think-f7fdc7059cf8
Nice overview on the actual issues with AI tech - and also some ideas on what to do with them. The idea of us having to rethink ethical values is especially strong. Though I think machine learning is pretty great showing us the difference between what we think about ethical values and what we actually do.
Link to post
Posted on 2017 December 17 by polgarp · Tagged in Children and Learning bias
Original link: https://magenta.as/the-new-kid-defense-the-algorithm-made-me-do-it-23cf7585e082
Scary results, but at the same time a little but funny. Algorithms are bound change how we interact with technology and this is especially true for children who may never met technology working differently. So maybe as we need different design for children and adults we may need different algorithms too.
Link to post
Posted on 2017 December 17 by polgarp · Tagged in Overview
Original link: https://www.uxmatters.com/mt/archives/2017/11/smartware-ai-and-magical-products.php
Mostly just an intro article into AI topics, but contains a nice description on how to think about AI tech. Rather just as something magical happening, it’s the same interactive system as non AI techs, it just does more based on the same user input.
Link to post
Posted on 2017 December 17 by polgarp · Tagged in Summary, Healthcare, Radiology, and Job automation
Original link: https://becominghuman.ai/radiologists-as-knowledge-experts-in-a-world-of-artificial-intelligence-summary-of-radiology-ec63a7002329
Radiology is on the fields in healthcare that comes up again and again as something relatively easy to automatize. So it’s interesting to see how radiologists see these developments, and how they would influence their future jobs.
Link to post
Posted on 2017 December 17 by polgarp · Tagged in Co-creation and Sci-fi
Original link: https://www.wired.com/2017/12/when-an-algorithm-helps-write-science-fiction/
Fascinating story of how a co-writing algorithm works embedded into a nice little sci-fi story. A few things to note here. First the initial data had some very unexpected results (like the number of women speakers). Second, and this kind of expected, some of the advices were highly statistical. For a better experience these need more semantics. Third the end result kinda lacked something, like a character. But still such applications might be good trainers - preparing creative artists for real works.
Link to post
Posted on 2017 December 17 by polgarp · Tagged in Human centered automation
Original link: https://uxdesign.cc/human-centered-automation-80890e1a764c
The idea of not automating every step of a workflow just because we can feels very useful. We should think about the whole system of human and algorithmic actors working together and optimize it for the humans involved not just for (potential) optimum reachable with automation.
Link to post
Posted on 2017 December 10 by polgarp · Tagged in Job automation and Data scientists
Original link: https://www.technologyreview.com/s/608921/you-could-become-an-ai-master-before-you-know-it-heres-how/
Very meta when you try to automatise the jobs of machine learning experts with machine learning. Also ironic that to atomatize more jobs we need to automatize jobs. When compared to the similar problems the UX field is facing I feel our best efforts to have more people doing design (bootcamps, design tools, off-the-shelf design systems, plugable user-insight generators) only resulted in the need for more designers. Hard to see how this will be different for data scientists.
Link to post
Posted on 2017 December 10 by polgarp · Tagged in AI design and Interaction
Original link: https://www.humanautonomy.com/blog/2017/11/3/throwback-thursday-ironies
Nice find on this article over 30 years ago, and it feels more relevant then ever. If we automatize all the easy things in a workflow, we leave only the hard things for the human operator. If we apply machine learning to all the things that we can easily collect data for, so humans can go for decisions on the strategic level, rather on the tactical level - we keep the humans from learning the tactics. Doesn’t sound like successful design on the long-term.
Link to post
Posted on 2017 December 10 by polgarp · Tagged in Data collection and Data security
Original link: https://9to5google.com/2017/12/05/ai-type-data-leak/
This is bad, especially for the people whos data was compromised. And overall something we can expect to happen more and more often. As AI tech needs more data about people, companies would want to hoard more and more data. It’s bound to happen these data vaults will leak. We designers like to think about ourselves that we deeply care about the people for whom we design for. To strike a balance between data collected and thinking about people’s data security seems like a nice a design challenge.
Link to post
Posted on 2017 December 10 by polgarp · Tagged in Apple, Research, and Data privacy
Original link: https://machinelearning.apple.com/2017/12/06/learning-with-privacy-at-scale.html
Apple’s promise of keeping the user’s privacy while still coming up with features powered by machine learning seems to provide a bunch of challenging problems - and this research sheds some light how are they actually doing it.
Link to post
Posted on 2017 December 10 by polgarp · Tagged in AI chips, Smartphone, and Honor
Original link: https://www.theverge.com/2017/12/5/16737446/honor-launch-flagship-view-10-7x-phone-animoji-face-id
AI chips are getting commoditized fast. Seems very soon having an AI chip won’t really differentiate a device, maybe just as the performance for learning applications. So there are a bunch of new design patterns to think about to display to the users when the learning within the app is slow.
Link to post
Posted on 2017 December 10 by polgarp · Tagged in DeepMind, Research, and Chess
Original link: https://www.theverge.com/2017/12/6/16741106/deepmind-ai-chess-alphazero-shogi-go
While this is a nice improvement over previous results altogether it’s not that suprising given how go is much more difficult as chess and just shows how far the field has come in recent years.
Link to post
Posted on 2017 December 9 by polgarp · Tagged in Intelligence augmentation and AI design
Original link: https://distill.pub/2017/aia/
Interesting longread about how to approach AI tech as human augmentation tool. There are not that many discussions on how humans’ thinking may change with such tools, so well worth the read.
Link to post
Posted on 2017 December 9 by polgarp · Tagged in Intercom and Interview
Original link: https://blog.intercom.com/intercom-fergal-reid-ai-machine-learning/
Nice interview with Intercom’s machine learning strategist.
Link to post
Posted on 2017 December 9 by polgarp · Tagged in Google, Co-creation, and Collecting data
Original link: https://www.blog.google/topics/machine-learning/quick-draw-one-billion-drawings-around-world/
Summary / update on the Quick, Draw! project. Besides a fun activity it also shows how data collection from users can be fun.
Link to post
Posted on 2017 December 4 by polgarp · Tagged in Interview and Overview
Original link: https://designmind.frogdesign.com/2017/10/current-state-intelligent-systems/
I like this interview with frog’s Sheldon Pacotti, especially the idea of the philosophical underpinnings of designing AI tech.
Link to post
Posted on 2017 December 4 by polgarp · Tagged in AI maturity and Strategy
Original link: https://hackernoon.com/the-ai-hierarchy-of-needs-18f111fcc007
This seems to be an excellent overview on the steps a company should take to be actually able to apply AI tech within their services or products. I’d even call it a maturity model, as you move forward you become more and more focused on applying certain processes and methods. Though it’s not stated explicitly I’d assume the transformation to be more effective with AI tech also covers organizational changes and even mindset changes.
Link to post
Posted on 2017 December 4 by polgarp · Tagged in Ethics, Policy, and Bias
Original link: https://www.nytimes.com/2017/12/02/opinion/sunday/intelligent-policing-and-my-innocent-children.html
Design AI tech based on past data is core to how these algorithms work. But doing this without thoroughly understand that data and the context around the data will result in bias and ultimately outcomes that were unintentional. Very similar to how user experience happens whether we apply research and design methods or no - only the outcomes will be vastly different.
Link to post
Posted on 2017 December 4 by polgarp · Tagged in Explainable AI and Overview
Original link: https://www.nytimes.com/2017/11/21/magazine/can-ai-be-taught-to-explain-itself.html
Fascinating article about the current state of explainable AI tech, and some of the implications.
Link to post
Posted on 2017 December 4 by polgarp · Tagged in Statistics and Overview
Original link: http://www.aiindex.org/
Nice report with lots of interesting data about AI tech, very useful when investigating trends.
Link to post
Posted on 2017 December 3 by polgarp · Tagged in Comics and Fun
Original link: http://www.smbc-comics.com/comic/ai-2
“How can a collection of wires and switches possibly experience true consciousness?”
Link to post
Posted on 2017 December 3 by polgarp · Tagged in Game design, Video games, AI creativity, and Co-creation
Original link: https://www.technologyreview.com/s/609482/ai-is-dreaming-up-new-kinds-of-video-games/
Just generating random concepts that might surprise a game designer or would be a new game is not much better than for example pulling out terms from a hat. What’s interesting here how the system goes beyond by understanding what would be considered as a good game and also generating the game itself - that would support the thinking process of the game designer.
Link to post
Posted on 2017 November 28 by polgarp · Tagged in Book and Ethics
Original link: https://www.npr.org/sections/alltechconsidered/2017/09/11/550177421/world-without-mind-how-tech-companies-pose-an-existential-threat
Looks like an interesting book on today’s tech challenges. I don’t think companies are necessarily evil but they do tend to have their own goals, most prominently profit. Much of how they reach it depends on designers. We already know interaction dark patterns, and they are well described, so it is easy to recognize them. The next step is to describe algorithmic dark patterns then.
Link to post
Posted on 2017 November 27 by polgarp · Tagged in AI design, Assistant, Co-creation, and Centaur design
Original link: https://www.fastcodesign.com/90146967/the-next-user-you-design-for-wont-be-a-human
“Centaur design” is a nice term to describe how to design products for humans and AI tech working together. It still seems difficult to articulate how such systems would work, and especially how to design for them.
Link to post
Posted on 2017 November 27 by polgarp · Tagged in Autonomous vehicles and User research
Original link: https://www.wired.com/story/mit-humans-semiautonomous-car-study/
Observing people interacting with technology is a common practice among software designers, so it makes sense a similar approach would help when working on self-driving cars. Just looking at data (logs from the system, the human operator’s actions) won’t give a full picture, especially not on the why. But to really understand the people, it’s not enough to silently observe them through cameras and sensors, questions need to be asked too.
Link to post
Posted on 2017 November 27 by polgarp · Tagged in Facebook and Suicide prevention
Original link: https://www.fastcompany.com/40498963/how-facebooks-ai-is-helping-save-suicidal-peoples-lives
Is this works, it would be pretty awesome. Since Facebook already has enough data to have an indication if somebody is suicidal and also enough capability to act on it. I wonder what other human issues could they detect.
Link to post
Posted on 2017 November 27 by polgarp · Tagged in Policy and Ethics
Original link: http://www.afr.com/technology/chief-scientist-alan-finkel-calls-for-artificial-intelligence-regulation-20171113-gzkt17
Asimov stamp for ethical or more generally a kind of certification for algorithmic products / services comes up many times. I guess mostly since similar things stamps worked earlier. The thing is - some of the bias can be decreased with the right approach, but what most people perceive as “bad AI” mostly comes from two things. One is malevolent actors, not sure a stamp would help on this. Second is that these type of algorithms however they are designed build on data and just amplify existing bias. Like a mirror, we are looking at ourselves.
Link to post
Posted on 2017 November 21 by polgarp · Tagged in Interview, Business model, and Product development
Original link: https://machinelearnings.co/why-ai-companies-cant-be-lean-startups-734a289792f5
Nice interview with Matt Turck of FirstMark Capital. Overall AI startups are very similar to other startups, only data and tech is not as easy yet as with other services. More like hardware based companies, as prototyping and getting things right takes exceptional talent and longer time. Matt didn’t mention design as an issue, most probably since these startups need the technology right first before they can invest more on their actual apps. But is it wise to invest into tech if you don’t understand your customers well enough? Leads to even more time prototyping.
Link to post
Posted on 2017 November 20 by polgarp · Tagged in Weaponized AI and Overview
Original link: https://gizmodo.com/hackers-have-already-started-to-weaponize-artificial-in-1797688425
Currently the most dangerous AI is not one that would destroy us on its own, but one that is controlled by malevolent humans. The title is a bit clickbait-y, but the threat seems real.
Link to post
Posted on 2017 November 20 by polgarp · Tagged in Robotics and Essay
Original link: http://rodneybrooks.com/forai-domo-arigato-mr-roboto/
Nice story on how robots help the clean up at Fukushima. And also about a limited AI they have - but one that works under real conditions.
Link to post
Posted on 2017 November 20 by polgarp · Tagged in Overview and Government
Original link: http://algorithmtips.org/
A cool list of algorithms used by the US government. Though some of it is just a simple formula and cannot be considered as an AI algorithm (like no training data), still quite a lot use existing data as a basis.
Link to post
Posted on 2017 November 20 by polgarp · Tagged in Overview
Original link: https://www.strategy-business.com/article/A-Strategists-Guide-to-Artificial-Intelligence
Nice overview on AI tech. I liked the separation of Assisted / Augmented / Autonomous Intelligence describing well the level of complexity of individual AI techs.
Link to post
Posted on 2017 November 19 by polgarp · Tagged in Emotional design and Psychology
Original link: https://medium.com/@alasaarela/the-human-api-f725191a32d8
A thought-provoking article introducing humans as we would have an API that can be driven by algorithms. Designers and products are already using this API in certain ways (think attention economy), but imaging basic human qualities as handles that can be pulled to arrive at a desired output feels like a pretty good description. Only now pulling is done by machines, not other people.
Link to post
Posted on 2017 November 19 by polgarp · Tagged in Talk and AI design
Original link: https://www.mindtheproduct.com/2017/11/product-design-era-algorithm-josh-clark/
Josh Clark has some of the best ideas in product design, so it’s well worth to listen to this talks about AI design.
Link to post
Posted on 2017 November 19 by polgarp · Tagged in Product, Lawyer, Assistant, and Bot
Original link: https://ndalynn.com/
“Hi! I’m Lynn, an AI lawyerbot who can review your NDA. I’ll tell you if it’s safe to sign on the dotted line.” Two interesting things here. First, as we get more and more such assistants helping people in everyday tasks that would otherwise need specialized knowledge, the less value it will have to create relationships between people relying on information asymmetry. Second, this bot is a marketing front for an IT lawyer. Just as apps were at a time marketing channels, so are now bots.
Link to post
Posted on 2017 November 19 by polgarp · Tagged in Apple, Face detection, and Computer vision
Original link: https://machinelearning.apple.com/2017/11/16/face-detection.html
A few interesting details - besides the Face ID’s working of course. Apple started to work on face detection in 2014. Should give an idea how long a truly innovative feature needs until it has sufficient polish.
Link to post
Posted on 2017 November 13 by polgarp · Tagged in Emotion, Emotional intelligence, and Dark AI
Original link: https://machinelearnings.co/the-rise-of-emotionally-intelligent-ai-fb9a814a630e
This article paints a scary image, AI being more emotionally intelligent, since it can understand emotions or even manipulate it more than most people even at it’s current state. That is, the companies deploying these algorithms getting better at knowing us and hijack our brains. And we were afraid of dark patterns in UX design.
Link to post
Posted on 2017 November 13 by polgarp · Tagged in Bot and Scam
Original link: https://www.rescam.org/
Wow. Pretty awesome webdesign too. “Introducing Re:scam – an artificially intelligent email bot made to reply to scam emails. Re:scam wastes scammers time with a never-ending series of questions and anecdotes so that scammers have less time to pursue real people.”
Link to post
Posted on 2017 November 13 by polgarp · Tagged in IBM, Watson, and Healthcare
Original link: https://www.statnews.com/2017/09/05/watson-ibm-cancer/
The Wizard of Oz method is one of the suggested ways of prototyping AI experiences, but that is reserved for the design phase, not for something after deployment. Besides the huge story of IBM’s marketing overhyping some things, there is an other moral here: product development with AI tech is hard, and having the right data is even harder.
Link to post
Posted on 2017 November 13 by polgarp · Tagged in Audible, Content consumption, and Reading
Original link: https://www.theverge.com/2017/11/1/16588974/audible-romance-novel-machine-learning-skip-to-the-good-part-audiobooks
This sounds like an interesting experiment. The most scarce resource we humans have is time, and it’s not just work that is ripe for automation, but maybe also how we spend parts of our free time.
Link to post
Posted on 2017 November 12 by polgarp · Tagged in Bias, Adversarial, Ethics, and Talk
Original link: https://points.datasociety.net/your-data-is-being-manipulated-a7e31a83577b
Bias is an important issue of anyone working with data and thus with AI tech, but it’s effects are sometimes not that understood. Besides creating awkward or plain bad product experiences. This post (and talk) sheds some light on potential dangers and consequences.
Link to post
Posted on 2017 November 12 by polgarp · Tagged in Assistant
Original link: https://chatbotsmagazine.com/the-5-levels-of-meeting-ai-b60187e51409
More like ideas of possible products how AI could be incorporated into meetings. The issue seems that it’s easy to imagine such products, but technological challenges aside how should be an AI assistant designed talks during meetings? The case presented (“Hey here is a report you might find interesting”) effect on the human participants is questionable I think. On a meeting not just the semantics need to be understood, but also the wider context: the organization (different cultures would need different AI personality), and the people’s relations to each other.
Link to post
Posted on 2017 November 12 by polgarp · Tagged in Overview and Design tools
Original link: https://uxdesign.cc/how-ai-will-impact-your-routine-as-a-designer-2773a4b1728c
Nice overview on the impact of AI tools on designers everyday life. The notion of AI tech doing much of the design legworks seems right - technology has been helping us for quite a while to focus on what matters and automatize everything else.
Link to post
Posted on 2017 November 12 by polgarp · Tagged in Talk, AI design, and Design research
Original link: https://www.youtube.com/watch?v=jLTdndKysN4
An interesting short talk bridges the deeper levels of both ML research and design research.
Link to post
Posted on 2017 November 7 by polgarp · Tagged in Content automation, Abuse, and YouTube
Original link: https://medium.com/@jamesbridle/something-is-wrong-on-the-internet-c39c471271d2
Very disturbing piece on what gets shown to young children on YouTube. I don’t see a clear evidence of AI tech in works, but based on the examples shown it could be very well some more sophisticated algorithms, designed to extract profit at scale. If this is so, this would be a very clear example how AI tech can be abused, as a kind of dark pattern to generate money while harming people.
Link to post
Posted on 2017 November 6 by polgarp · Tagged in Data bias
Original link: http://www.slate.com/articles/technology/technology/2017/10/what_happens_when_the_data_used_to_train_a_i_is_biased_and_old.html
It’s important to realize that most publicly available, and also a huge portion of non-public non-trivial data sets will contain significant bias that distorts the results the algorithms. The simple fact is, earlier datasets were not collected according to today’s standards on bias, or not created with avoiding bias in mind. Since data is needed to design (maybe even just to prototype) and release any type of AI tech based product this should be the concern for most product teams. I’m not sure if this type of bias can be uncovered by quantitative analysis. User research should help by understanding the human context around the data.
Link to post
Posted on 2017 November 6 by polgarp · Tagged in Fun, Comics, and Computer vision
Original link: http://www.commitstrip.com/en/2017/11/03/vision-api-fail/
Cute muffins, eh?
Link to post
Posted on 2017 November 6 by polgarp · Tagged in Airbnb and Design tool
Original link: https://airbnb.design/sketching-interfaces/
Turning sketches into designs and code, even when built on a design system is quite awesome. I like how this gives a glimpse how UX design may happen in the future. Today’s designers will be system designers, and anybody can create a specific instance of design assisted by AI tech.
Link to post
Posted on 2017 November 6 by polgarp · Tagged in Research and Adversarial attacks
Original link: http://www.labsix.org/physical-objects-that-fool-neural-nets/
This is troubling research. Most AI tech applications can be designed with well-intentioned users in mind, if an error happens, no harm done. But as we replace more and more safety critical applications with AI tech, such attacks may fool systems designed to have better performance than humans.
Link to post
Posted on 2017 November 6 by polgarp · Tagged in Co-creation, Writing, Fun, and Product
Original link: https://cyborg.tenso.rs/
More fun than useful, this app actually showcases pretty well how everyday, mundane tasks (like writing a letter) could be augmented with AI tech.
Link to post
Posted on 2017 October 26 by polgarp · Tagged in Andrew Ng, Chatbot, and Mental health
Original link: https://medium.com/@andrewng/woebot-ai-for-mental-health-d0e8632b82ba
Andrew Ng joining a startup’s board to support them in building a depression fighting chatbot. This is excellent news for anybody concerned on mental health.
Link to post
Posted on 2017 October 26 by polgarp · Tagged in AI design and Ideo
Original link: https://www.fastcodesign.com/90147010/exclusive-ideos-plan-to-stage-an-ai-revolution
Since Ideo is one of the leading design firms in the world, it’s well worth a look when they do something. The acquisition of a data science firm seems like a great mix to create awesome, human-centered experiences with AI tech.
Link to post
Posted on 2017 October 25 by polgarp · Tagged in Assistant and Online dating
Original link: https://gizmodo.com/the-future-of-online-dating-is-unsexy-and-brutally-effe-1819781116
I wonder how well past behavior can actually predict the success of date or even a relationship. As also the article mentions, date outcomes would be the most important data to look for. Also people outsourcing there sense of an other person (love at first sight?) to AI-based assistants poses interesting questions regarding what type of experiences would this result in on the long term?
Link to post
Posted on 2017 October 25 by polgarp · Tagged in Policy and Ethics
Original link: https://medium.com/@AINowInstitute/the-10-top-recommendations-for-the-ai-field-in-2017-b3253624a7
Nice set of recommendations, with missing policies I wonder why companies would choose to follow these and what would be some of the impact on product development. The comparison to drug trials is interesting, especially that considering the scale AI tech could have a much bigger impact on people’s life.
Link to post
Posted on 2017 October 25 by polgarp · Tagged in AI design, Recall, and Precision
Original link: https://medium.com/@matyas_82059/recall-vs-precision-in-an-ai-product-or-feature-cc3f1f97175a
Recall and precision are important concepts to balance when designing a system with AI tech, this article gives a good overview on them. I liked the last part explaining that finally even this decision doesn’t need to be exclusive, it can be a fluid depending on the available data.
Link to post
Posted on 2017 October 25 by polgarp · Tagged in Product and Privacy
Original link: https://www.theverge.com/2017/10/17/16414822/nude-app-photo-vault-sexting
Feels like a nice app. One, machine learning is used for a feature that wouldn’t have been possible without, the convenience of not having to look through thousands of images in one’s library. Two, it reflects on something very basic people are doing instead of shy assumptions - people taking nude photos of themselves for whatever reasons.
Link to post
Posted on 2017 October 25 by polgarp · Tagged in Facebook and Translation
Original link: https://www.theverge.com/us-world/2017/10/24/16533496/facebook-apology-wrong-translation-palestinian-arrested-post-good-morning
Translations, especially translation mistakes can have grave impact on real life, like this case shows. Makes you wonder what error types should you allow in translations - and also if it might be a good idea to design in a way that signals confidence. Humans speaking an other language have a good grasp on how precise is the meaning they got from the context.
Link to post
Posted on 2017 October 25 by polgarp · Tagged in Assistant and Overview
Original link: https://www.recode.net/2017/10/18/16492156/coaching-cloud-future-work-jobs-artificial-intelligence-ai-enterprise-employee-training
Even if it sounds a little bit like a made up marketing world, I like the expression “coaching cloud”. It well summarizes how AI tech can be deployed in a human-centric way, by enabling and supporting people to get better at what they do over time. This is something we do any way when we analyze case studies and industry reports, and well defined methods can use scale to improve even nuanced details.
Link to post
Posted on 2017 October 25 by polgarp · Tagged in Overview, AI chips, and Mobile
Original link: https://www.theverge.com/2017/10/19/16502538/mobile-ai-chips-apple-google-huawei-qualcomm
Nice overview on the current state of mobile “AI chips”. While they are currently mostly for flagship devices, soon they’ll be just another component in the SoC of the phone.
Link to post
Posted on 2017 October 23 by polgarp · Tagged in Ethics and Bias
Original link: https://theoutline.com/post/2401/what-would-the-average-human-do
Setting the course for ethical choices of an AI via democracy sounds like a good approach - as we are already doing this in the form of policy making (most countries anyway). But the scale and AI’s sometimes surprising behavior can lead to unexpected results. Designing products with crowdsourced ethics will be an interesting challenge too, as these decisions rested sofar on the designers.
Link to post
Posted on 2017 October 23 by polgarp · Tagged in AI jobs
Original link: https://www.nytimes.com/2017/10/22/technology/artificial-intelligence-experts-salaries.html
Just the sign of times, AI specialists get salaries like rock stars.
Link to post
Posted on 2017 October 23 by polgarp · Tagged in Google, Research, Photography, and Computational photography
Original link: https://research.googleblog.com/2017/10/portrait-mode-on-pixel-2-and-pixel-2-xl.html
A quite detailed description on the Portrait Mode feature, also a nice example how machine learning can be embedded into a product pipeline. We hear it many times that software is eating the world, well here it eats part of the appeal of SLR cameras.
Link to post
Posted on 2017 October 23 by polgarp · Tagged in Statistics
Original link: https://srconstantin.wordpress.com/2017/01/28/performance-trends-in-ai/
If you are wondering how AI currently performs over various domains, this article contains tons of statistics. What’s especially interesting is the comparison to human performance for example in speech or image recognition. But keep in mind, as algorithms get better than humans, they also make different kind of mistakes.
Link to post
Posted on 2017 October 23 by polgarp · Tagged in Overview, Research, and Music generation
Original link: https://medium.com/artists-and-machine-intelligence/neural-nets-for-generating-music-f46dffac21c0
Nice overview on the current state of algorithmic music creation with neural nets with many examples.
Link to post
Posted on 2017 October 23 by polgarp · Tagged in DeepMind, Go, and Research
Original link: https://deepmind.com/blog/alphago-zero-learning-scratch/
This is pretty exciting, as this version doesn’t build on existing data, it just knows the initial rules of Go and plays lots of games knowing these rules.
Link to post
Posted on 2017 October 17 by polgarp · Tagged in Chatbots and Conversational UI
Original link: https://www.linkedin.com/pulse/most-chatbots-dont-use-ai-misrepresenting-riza-c-berkan-ph-d/
Subject says it all. Chatbots may be a good approximation for some AI tech related experiences, probably that is why some people confuse the two. But most chatbots are just a conversational UI with a few IFs behind. Good for prototyping the experience for a real AI system.
Link to post
Posted on 2017 October 17 by polgarp · Tagged in Future AI and Consciousness
Original link: https://www.wired.com/story/how-to-build-a-self-conscious-ai-machine/
Fascinating long read about the human consciousness and how might we build a self-conscious machine.
Link to post
Posted on 2017 October 17 by polgarp · Tagged in Bias, AI apocalypse, and Machine learning
Original link: https://www.technologyreview.com/s/608986/forget-killer-robotsbias-is-the-real-ai-danger/
If your data is biased, your algorithms will be biased and ultimately your product will be flawed. Open training data, or at least some published metrics about it might be an important way assessing future products. Should this be part of a designers job, being sure that the data doesn’t contain biases? Especially when talking about people’s data, design research might be equipped for this. Or will be this a specialized job, something mixed between data science and user research?
Link to post
Posted on 2017 October 17 by polgarp · Tagged in Facebook, AI design, Use case, and Product development
Original link: https://medium.com/elegant-tools/designing-with-ai-3f7652619f4
Excellent use case on how Facebook’s Recommendations product included AI tech during their design process. Especially interesting are the details about how they ended up with using AI tech - initially they just had a problem to solve.
Link to post
Posted on 2017 October 17 by polgarp · Tagged in Google, Product, and AI chips
Original link: https://www.theverge.com/2017/10/4/16405200/google-clips-camera-ai-photos-video-hands-on-wi-fi-direct
A very interesting product with a few nice design tips for similar AI tech based products, like have a button to provide agency. Unlike Google that Clips has AI on chip instead in the cloud, this makes it compelling to use even for privacy concerned users. These just show how much thinking on different aspects of the design went into it.
Link to post
Posted on 2017 October 17 by polgarp · Tagged in Slack and Product
Original link: https://www.technologyreview.com/s/608953/slack-ceo-how-well-use-ai-to-reduce-information-overload/
Interesting to see how Slack intends to apply AI tech. Improving internal communication is essential especially for larger organizations. Ofcourse this kind of implies that most communications happen online which is not always the case. Would people chat via Slack more, if they would get feedback on their style?
Link to post
Posted on 2017 October 16 by polgarp · Tagged in Ethics and DeepMind
Original link: https://deepmind.com/blog/why-we-launched-deepmind-ethics-society/
Ethics with AI tech is an important topic, and there are many articles about how to progress on this field. This it’s a great step forward when one of the foremost research organization on this field decides to dedicate resources.
Link to post
Posted on 2017 October 16 by polgarp · Tagged in Fun, Video game, and Toy
Original link: http://www.decisionproblem.com/paperclips/index2.html
Fun little clicker game based on the premise of the paperclip producing AI. A variation of the AI apocalypse theme, what happens if an AI’s goal would be producing paperclips, not caring about humanity and the planet.
Link to post
Posted on 2017 October 16 by polgarp · Tagged in Research and IQ
Original link: https://arxiv.org/abs/1709.10242#
Grading the IQ of AI systems seems like an interesting experiment, especially to compare the “performance” of different company. Not sure though if the relation to human intelligence means anything, increasing machine intelligence doesn’t seem like a linear process. Also highly intelligent systems might appear dumb to a casual observer, as it may lack emotional value, really depends on the design.
Link to post
Posted on 2017 October 11 by polgarp · Tagged in Google, Experiment, and Toy
Original link: https://teachablemachine.withgoogle.com/
I was writing just the other day about how difficult is to play with AI technologies, something that makes it hard for designers to get a feel for the “AI material”. Now this experiment is created for children, still it offers a nice glimpse into what is possible. Well worth to try it.
Link to post
Posted on 2017 October 11 by polgarp · Tagged in Microsoft and Overview
Original link: https://www.linkedin.com/pulse/how-microsoft-uses-artificial-intelligence-drive-business-marr/
A little bit PR smelling, but still this article provides a nice overview on Microsoft’s effort to make a dent in the AI field. They always had strong research people, but it seems Microsoft has a difficulty to translate these results into well designed product. Maybe their ongoing transformation will change this.
Link to post
Posted on 2017 October 11 by polgarp · Tagged in China and Overview
Original link: https://www.technologyreview.com/s/609038/chinas-ai-awakening/
Pretty nice overview on AI tech efforts in Chine. For all the talk about a new “space race” focused on AI, there is definitely something happening there.
Link to post
Posted on 2017 October 11 by polgarp · Tagged in Analysis and Quantum computing
Original link: http://www.macdrifter.com/2017/10/artificial-intelligence-super-computers-and-society.html
A nice, thoughtful opinion piece on the current direction of AI technology and it’s impact on society. Especially the part on quantum computing is interesting, it does sound like something that would raise the performance of AI tech significantly.
Link to post
Posted on 2017 October 8 by polgarp · Tagged in Design education and AI design
Original link: https://www.fastcodesign.com/90145027/designers-arent-equipped-to-make-ai-heres-how-to-prepare
The chief issue seems to be that AI tech is hard to play with, as results are unpredictable sometimes. Hopefully better tools will be available in the near future making this easier. The other is while designers work now pretty well with engineers, data scientist are different. We need to learn to work together, probably the best starting point is storytelling with data.
Link to post
Posted on 2017 October 8 by polgarp · Tagged in Google, Talk, AI design, Overview, and Machine learning
Original link: https://www.youtube.com/watch?v=FpgbqgAEPfo
Pretty nice talk from googlers on their take on designing systems with machine learning. Many interesting details, especially their focus on biases and getting it right.
Link to post
Posted on 2017 October 8 by polgarp · Tagged in AI design and Design challenges
Original link: https://www.fastcodesign.com/90144510/3-problems-with-ai-only-designers-can-solve
I like all three of these problems, as they relate pretty well to non AI related product problems: make the system’s state clear, choose your user problem to focus on before technology, and involve designers within product teams from day one. The last one feels the most challenging to actually pull through, as design in most companies has still a shallow impact - only user related functions. But everybody needs to understand, whenever there is a decision that impacts the user in any way (like database speed), there needs to be somebody who explicitly champions the user needs. Designers seem to be well suited for this, but their understanding of their role needs to change too.
Link to post
Posted on 2017 October 7 by polgarp · Tagged in Design tools, Talk, and Co-creation
Original link: https://www.youtube.com/watch?v=aR5N2Jl8k14
Nice talk about the potentials of generative design tools.
Link to post
Posted on 2017 October 7 by polgarp · Tagged in Future AI, Long read, and Overview
Original link: https://rodneybrooks.com/the-seven-deadly-sins-of-predicting-the-future-of-ai/
Insightful and also entertaining long read about future AI. Or really about what most people believe, current AI. Many examples and analogies from other fields, it also provides some ideas on challenges for AI powered products, like deployment difficulties.
Link to post
Posted on 2017 October 7 by polgarp · Tagged in Research, Future AI, and Deep learning
Original link: https://www.technologyreview.com/s/608911/is-ai-riding-a-one-trick-pony/
Fascinating read on the current thinking on AI tech, or more precisely on deep learning. I also used the term “AI revolution” a few times, but that also seems a matter of perspective for answering people’s needs even the current level of narrow AI opens up many new opportunities.
Link to post
Posted on 2017 October 7 by polgarp · Tagged in Overview and Fintech
Original link: https://medium.com/towards-data-science/artificial-intelligence-verticals-ii-fintech-5dcffae10b1
Nice overview on where fintech stands with the adoption of AI tech with also a rationale of how these technologies are introduced. Very useful if you are looking to get started with a fintech product.
Link to post
Posted on 2017 October 2 by polgarp · Tagged in TED talk, Talk, Research, and Robotics
Original link: https://www.ted.com/talks/radhika_nagpal_what_intelligent_machines_can_learn_from_a_school_of_fish
Pretty nice talk on simple robots exhibiting collective intelligence. Although these are very simple use cases, it’s easy to imagine how smarter machines would show much more complex interactions together. Combining this with the last idea of what patterns human behavior exhibits leads to some interesting thought experiments.
Link to post
Posted on 2017 October 2 by polgarp · Tagged in AI design and Human-centered machine learning
Original link: https://becominghuman.ai/ux-design-for-implicit-and-explicit-feedback-in-an-ai-product-9497dce737ea
One of the more interesting part of designing AI experiences is how the humans can interact not only with the results of an AI algorithm, but also how those results are created. This article explores feedback as a control mechanism.
Link to post
Posted on 2017 October 2 by polgarp · Tagged in Job automation and Principles
Original link: https://boingboing.net/2017/09/27/robots-stealing-jerbs.html
Nice set of rules, and of course they could be applied to “non-physical robots”, so algorithms. These laws (?) actually make nice design constraints too, and maybe even can act as principles on what design and especially research efforts should go into such a system.
Link to post
Posted on 2017 October 2 by polgarp · Tagged in AI design tool and Product
Original link: http://khroma.co/
Feels a little bit basic, but nonetheless the results are quite intriguing. One more tool to take out some tedious work from a designer’s day.
Link to post
Posted on 2017 October 2 by polgarp · Tagged in Google, Overview, and Journalism
Original link: https://www.blog.google/topics/journalism-news/how-publishers-can-take-advantage-machine-learning/
Nice overview on Google tools supporting journalists. Looking beyond the basic advertisement style, most jobs could make a similar inventory as we use more and more tools with AI tech.
Link to post
Posted on 2017 October 2 by polgarp · Tagged in Google and AI apocalypse
Original link: https://techcrunch.com/2017/09/19/googles-ai-chief-thinks-reports-of-the-ai-apocalypse-are-greatly-exaggerated/
We need a better definition for the AI apocalypse. We probably won’t see T-800s coming out of factories. We already see AI tech used in scaling up spreading fake news and personalized propaganda. I wonder what Giannandrea thinks about that.
Link to post
Posted on 2017 October 2 by polgarp · Tagged in Facebook
Original link: https://www.recode.net/2017/9/14/16310512/facebook-mark-zuckerberg-algorithm-ad-targeting-jews
Maybe not exactly an issue with their AI systems, but definitely has to do how Facebook creates their algorithms. Unfortunately such issues are not mistakes or errors within the algorithm, but more a problem on how these algorithms were designed. Just shows that (service) design needs to go much wider on this scale.
Link to post
Posted on 2017 October 2 by polgarp · Tagged in Machine learning and Resource
Original link: https://ml-showcase.com/
Nice list of ML projects, should be good for inspiration. Since they are also open source also a great list for learning.
Link to post
Posted on 2017 September 25 by polgarp · Tagged in Apple and AI chips
Original link: https://www.theverge.com/2017/9/13/16300464/apple-iphone-x-ai-neural-engine
Even though some people are saying this, Apple doesn’t seem to be behind the AI race. In contrary they now have machine learning capabilities onboard their latest phones, while also taking a different approach than most other (cloud?) companies. As for application / service design this means on hand iPhones allow for all kinds of nice features - and makes cross-platform development also harder.
Link to post
Posted on 2017 September 25 by polgarp · Tagged in Google and Research
Original link: https://research.googleblog.com/2017/09/the-google-brain-teams-approach-to.html
Nice overview how Google conducts basic research on machine learning. Also this seems like a working model of how to create an “innovation lab”, but that maybe only works at this scale. Especially awesome how they try to bring their results closer to non-researchers, like with the PAIR initiative and TensorFlow Playground.
Link to post
Posted on 2017 September 25 by polgarp · Tagged in Unity and Video games
Original link: https://www.fastcompany.com/40469609/machine-learning-is-making-video-game-characters-smarter-and-robots-more-competent
Unity having built in machine learning capabilities sounds like very cool. Robotics implications aside this opens up plenty of new opportunities for game designers. Also video games seem to drive many innovations, if nothing else, GPUs come to mind that enabled the current generation of AI tech.
Link to post
Posted on 2017 September 25 by polgarp · Tagged in Hate speech
Original link: https://www.fastcompany.com/40463759/why-alphabets-ai-has-such-a-hard-time-flagging-the-internets-worst-trolls?utm_source=feedly&utm_medium=webfeeds
Based on the contents, the title seems optimistic.
Link to post
Posted on 2017 September 24 by polgarp · Tagged in AI investment and AI race
Original link: https://medium.com/intuitionmachine/the-deep-learning-sputnik-moment-3e5e7c41c5dd
No wonder there are so many comparing the space race to the rate of investment organizations and nations are putting into AI. There are differences though, most people don’t understand the power of this technology, and how it’s already powering lots of parts of everyday life. The second is that there are no clear visible goals, like that of reaching the Moon that is also easy to understand by anybody. However it’s troubling that the best machine learning researchers in the west now work for corporations, less scientific breakthrough will have an overall impact on humanity.
Link to post
Posted on 2017 September 24 by polgarp · Tagged in Bias, Weapons of math destruction, and Mathwashing
Original link: https://www.mathwashing.com/
Mathwashing is a pretty awesome term to describe how algorithms (and thus AI techs) inherently carry the bias of their designers, this expecting neutral and rational behavior is s mistake. Besides a nice explanation the site provides links to additional info on this topic.
Link to post
Posted on 2017 September 24 by polgarp · Tagged in AI mistake and AI surprise
Original link: https://www.forbes.com/sites/kashmirhill/2012/02/16/how-target-figured-out-a-teen-girl-was-pregnant-before-her-father-did/#1fdcb4866668
Interesting not because of Target’s approach (should be pretty much standard by now), but the funny - and also somehow cruel anecdote within. The creepiness feeling people feel is one type of AI challenge, surprising behavior. That is the algorithm worked as intended, but it did cause surprise for humans who used and not really in a pleasant way.
Link to post
Posted on 2017 September 24 by polgarp · Tagged in AI policy, Ethics, and Face recognition
Original link: https://medium.com/intuitionmachine/high-time-to-begin-regulation-of-face-recognition-a-i-f4a92ee40165
I’m not sure regulation alone would solve using face recognition for the wrong purposes. But if it’s against the law, face recognition would be harder to use in a risky way. That is risky for the humans whose face was involved. For example having such systems as opt-in as default feels like a nice start (similar to how email is regulated). Opt-ins would be then a design challenge, how to convince people to allow their faces to be recognized?
Link to post
Posted on 2017 September 24 by polgarp · Tagged in Emotional design, Emotional intelligence, Emotions, and Empathy
Original link: http://consciousmagazine.co/artificial-intelligence-age-empathy/
Working towards an AI that also incorporates emotional intelligence is an interesting idea, especially as it strives to humanize and make AI more compatible to their human counterpart. Empathy as a vessel for emotions is nice first step for this. Although empathy in design feels like an overloaded concept, as it’s one of design’s core tools designers seem to be well equipped to lead these efforts.
Link to post
Posted on 2017 September 24 by polgarp · Tagged in Fun and Machine learning
Original link: http://lewisandquark.tumblr.com/post/165373096197/a-neural-network-learns-to-create-better-dd
“Aside from a couple of spells that just might work, most of the list is magicky-sounding nonsense, sometimes barely pronounceable.” Feels like a realistic list of spells students of Magic 101 would create.
Link to post
Posted on 2017 September 23 by polgarp · Tagged in OpenAI, Video games, Dota 2, and Research
Original link: https://blog.openai.com/more-on-dota-2/
Even more interesting info on the Dota 2 bot. The descriptions of the different matches almost feel like a coach talking about their team.
Link to post
Posted on 2017 September 23 by polgarp · Tagged in Policy and Ethics
Original link: https://www.nytimes.com/2017/09/01/opinion/artificial-intelligence-regulations-rules.html
Interesting proposal, especially comparing to Asimov’s laws. The thing is with the laws of robotics as Asimov described would be defined in the robots’ brains by default, there was no way to create a robot without the laws. I wonder how policy can be defined to have a such strong control over an AI. In any case these rules make sense and would drive AI developments into a more ethical direction.
Link to post
Posted on 2017 September 23 by polgarp · Tagged in Journalism and AI weapons
Original link: http://idlewords.com/2017/09/anatomy_of_a_moral_panic.htm
We will see more posts like this in the future, as we are not yet on peak-AI. Mainstream journalists misunderstanding how machine learning works, and what is it capable of (besides other issues in the story, like what would be a good and cheap way to blow things up). The problem seems to be, that compared to other technologies, AI is conceptually just more complicated and some concepts are hard to describe in layman’s terms, especially regarding scale.
Link to post
Posted on 2017 September 22 by polgarp · Tagged in Product design and Overview
Original link: https://medium.com/@polgarp/designing-ai-products-b66b1164634e
I did a talk on AI design at the UXBP meetup, and wrote up my notes into an article. Should be a good overview for designers wondering how to get started with product design for AI with plenty of examples and links.
Link to post
Posted on 2017 September 20 by polgarp · Tagged in Chatbots and Fun
Original link: https://chatbot.fail/
Fun collection of current chatbot fails, and also a little bit of depth on what are some of the reasons for it.
Link to post
Posted on 2017 September 20 by polgarp · Tagged in AI development and Bias
Original link: http://www.techrepublic.com/article/top-10-ai-failures-of-2016/
AI is not the silver bullet even when it’s perfectly, and it can be downright funny / scary when it’s poorly executed. Most of these fails relate to bias, but there are other issues too.
Link to post
Posted on 2017 September 17 by polgarp · Tagged in Microsoft, Chatbots, and Emotional computing
Original link: https://nautil.us/issue/52/the-hive/your-next-new-best-friend-might-be-a-robot-rp
Interesting read on a celebrity chatbot. There many nice details, like what design principles make the bot better conversation partner, and how they measure their success. One fear would be with an all-powerful, better than your human friends chat partner would be that we end up talking with each other even less. So I wonder, how such a technology could be used to have people talk more to each other.
Link to post
Posted on 2017 September 17 by polgarp · Tagged in Surveillance, Computer vision, and AI weapons
Original link: https://qz.com/1069756/using-artificial-intelligence-to-identify-protestors-wearing-hats-or-scarves-is-entirely-possible/
I agree, “Too many worry about what AI—as if some independent entity—will do to us. Too few people worry what power will do with AI.”.
Link to post
Posted on 2017 September 17 by polgarp · Tagged in Fake news, AI propaganda, and AI weapons
Original link: https://scout.ai/story/the-rise-of-the-weaponized-ai-propaganda-machine
Whenever I hear sentences about how AI gonna destroy the world, I don’t think about terminators killing people, but more about the events described in this article. Personalized messaging influencing people’s thinking and hiding reality. But just as AI can be used to spread fake news, it can also help people stay informed, and maybe even guard us from such propaganda.
Link to post
Posted on 2017 September 17 by polgarp · Tagged in Andrew Ng, Overview, and Machine learning
Original link: https://www.quora.com/session/Andrew-Ng/2
Answers on Quora to popular machine learning related questions by Andrew Ng. If you are looking for thoughtful, well informed answers on your frequent ML questions, this is the place to start.
Link to post
Posted on 2017 September 17 by polgarp · Tagged in Chatbots, Design principles, Human-computer interaction, and Politeness
Original link: https://chatbotsmagazine.com/rethinking-politeness-in-chatbots-interesting-research-findings-6c3a91871e5f
Humans attributing human-like features to things, so anthropomorphizing is nothing new, but it’s interesting to apply it as a design principle to systems with AI tech. Since AI can respond in unpredictable ways (that is, unpredictable in design time), they may appear even more human like. This possible also lowers user anxiety when working with automagical systems.
Link to post
Posted on 2017 September 17 by polgarp · Tagged in Comics, Fun, and SMBC
Original link: http://www.smbc-comics.com/comic/likely-apocalypse
Posted on 2017 September 17 by polgarp · Tagged in OpenAI, Research, and Theory of mind
Original link: https://blog.openai.com/learning-to-model-other-minds/
Fascinating research. Theory of mind is such an important factor of why humans are able to work together, having AI systems with the same attribute is huge step forward for both AI-AI and AI-human interaction.
Link to post
Posted on 2017 September 17 by polgarp · Tagged in AI design job and x.ai
Original link: https://x.ai/jobs-of-the-future-ai-interaction-designer/
Nice article on how x.ai’s designer works. Offers also a glimpse in the future of (UX?) design. Focusing on the user, the user flow, understanding both macro and micro views, connecting to code and applying psychological knowledge remains the same, but the tools and the medium might differ.
Link to post
Posted on 2017 September 17 by polgarp · Tagged in Microsoft, Education, and Product
Original link: https://play.google.com/store/apps/details?id=com.microsoft.mtutorclientandroidspokenenglish&hl=en
Pronunciation is one of the harder things in language learning, both to actually learn it, both to design it. Also not just “perfection” is acceptable, but “understandable” too. An ML based approach seems nicely suited for this.
Link to post
Posted on 2017 September 17 by polgarp · Tagged in Comics, Fun, and SMBC
Original link: http://www.smbc-comics.com/comic/ai
Posted on 2017 September 17 by polgarp · Tagged in Overview
Original link: https://monkeylearn.com/blog/machine-learning-on-the-customer-journey/
Nice overview on tools and frameworks to be used at different points of the customer lifecycle.
Link to post
Posted on 2017 September 16 by polgarp · Tagged in Google, Talk, and Overview
Original link: https://blog.ycombinator.com/jeff-deans-lecture-for-yc-ai/
Very nice talk on Google’s machine learning efforts, presented by Google Brain’s leader.
Link to post
Posted on 2017 September 12 by polgarp · Tagged in Design tool, Logo, and Graphic design
Original link: http://brandmark.io/logo-crunch/
One more tedious design task solved by a nice AI based tool: making high-res logos legible on lower resolutions.
Link to post
Posted on 2017 September 11 by polgarp · Tagged in AI-first
Original link: https://www.linkedin.com/pulse/new-rules-build-vs-buy-ai-first-world-michael-vaccarino
I don’t see build or buy such a clear cut question. Ofcourse if a company doesn’t have any technological capabilities to speak of they are better of buying a service. But in almost all the cases, the buying of AI technologies is risky because of bias. It’s not just data that introduces bias into AI, also the specific selection of all the parts and if a company doesn’t own its own stack, they risk unknown biases entering their systems. Such biases doesn’t exist with non-AI tech.
Link to post
Posted on 2017 September 11 by polgarp · Tagged in Overview, Design principles, and Interactive machine learning
Original link: https://medium.com/@atduskgreg/power-to-the-people-how-one-unknown-group-of-researchers-holds-the-key-to-using-ai-to-solve-real-cc9e75b1f334
Great overview and some nice principles to use when designing systems using machine learning. Greg also collected an awesome list of further reads on the topic of Interactive Machine Learning.
Link to post
Posted on 2017 September 11 by polgarp · Tagged in Product and Education
Original link: https://getparla.com/
Nice to see products appearing who claim to use AI technologies, but don’t make this the center of their message. Just solving a very human problem (learning a language) in personalized and effective way.
Link to post
Posted on 2017 September 11 by polgarp · Tagged in Music and Co-creation
Original link: https://github.com/schollz/PIanoAI
Awesome project that enables a piano player to get help from a piece of AI tech that plays the piano in the same style. If you have a co-player who is much better then you, say a computer trained on the best players’ work, you might get easily frustrated it doesn’t help to work in your own style.
Link to post
Posted on 2017 September 11 by polgarp · Tagged in Amazon, Microsoft, and Voice UI
Original link: https://www.fastcodesign.com/90138636/amazon-and-microsoft-just-made-a-major-bet-on-user-friendly-design
No, actually they didn’t. In theory connecting Alexa and Cortana might sound like a good idea, there are too many walled gardens after all. But this will complicate the clunky voice UIs in these product - for starters how will the users know whom do they talk to at any given time? This is not user-friendly design at all.
Link to post
Posted on 2017 September 11 by polgarp · Tagged in Ethics
Original link: https://thedecisionlab.com/magna-carta-inclusivity-fairness-global-ai-economy/
Magna Carta is a nice example where participants got more rights, and it seems to be beneficial to have something involving current participants in the field preventing short-sighted policy making. As AI tech can easily enable to exploit people for profit and even as a weapon this feels quite urgent.
Link to post
Posted on 2017 September 10 by polgarp · Tagged in Autonomous vehicles and Opinion
Original link: https://www.recode.net/2017/9/5/16257314/stop-autonomous-self-driving-cars-not-coming-soon-future
Nice overview on the current challenges of self-driving cars. The article misses though the point how the change will affect drivers, and if people are prepared for the change. Even the assisted driving needs different effort - would you need a different type of driver id for these types of car?
Link to post
Posted on 2017 September 10 by polgarp · Tagged in Product, Education, and Writing
Original link: https://www.fastcompany.com/40458272/this-machine-learning-powered-software-teaches-kids-to-be-better-writers
Posted on 2017 September 10 by polgarp · Tagged in AI race
Original link: https://www.theverge.com/2017/9/4/16251226/russia-ai-putin-rule-the-world
Feels like the space race all over again - major states are scrambling to get a technological advantage. Only this time the goals are less clear.
Link to post
Posted on 2017 September 10 by polgarp · Tagged in Apple, Photo manipulation, Product, and Design tool
Original link: https://www.theverge.com/2017/9/5/16254300/pixelmator-pro-features-launch-date
Looks amazing, another nice tool in a designer’s toolbox on Mac. What is more interesting how Pixelmator so extensively builds on Apple’s Core ML. This framework might mean, that even more smart apps will be Mac only - at least until Microsoft does something.
Link to post
Posted on 2017 September 10 by polgarp · Tagged in Agriculture and Acquisition
Original link: https://www.theverge.com/2017/9/7/16267962/automated-farming-john-deere-buys-blue-river-technology
Farmers can get all kinds of interesting assistance from AI tech, so much of their work relies on tedious tasks. So this is probably just one of the long line of purchases agriculture companies are making among AI startups.
Link to post
Posted on 2017 September 10 by polgarp · Tagged in Research, Online review, and Attacks
Original link: https://arxiv.org/pdf/1708.08151.pdf
One fascinating thing with AI tech is it’s ability to bust existing assumptions on what can and cannot be done in designing a product. This research shows that relying on user generated data, like reviews towards on outcome, like a nice rating for a product can be gamed.
Link to post
Posted on 2017 September 10 by polgarp · Tagged in Toyota, Manufacturing, and Job automation
Original link: https://www.fastcompany.com/40461624/how-toyota-is-putting-humans-first-in-an-era-of-increasing-automation
Fascinating read on Toyota’s recent changes to their manufacturing process especially the application of lean to using robots. Many interesting points here, like how people improving their own work processes are much like designers. But the most interesting part is how Toyota keeping the knowledge of handcrafting with their workers. This eliminates a huge concern in job automation, that newly appointed robot overseers don’t understand what their robots do, if they didn’t learn the craft by themselves.
Link to post
Posted on 2017 September 10 by polgarp · Tagged in Amazon, Robots, and Job automation
Original link: https://www.fastcompany.com/40463313/amazon-promises-good-human-robot-working-relationships-at-first-nyc-fulfillment-center
Fun title, as if humans will have to work alongside aliens. I wonder how such systems, that is humans and robots working together can be designed. Robots cannot do kaizen. Yet.
Link to post
Posted on 2017 September 10 by polgarp · Tagged in Video games and Research
Original link: https://www.theverge.com/2017/9/10/16276528/ai-video-games-game-engine
Playing a game is much easier then being able to describe how the game works - and this is true for using all other (non-game) systems.
Link to post
Posted on 2017 September 7 by polgarp · Tagged in Google, Youtube, Personalization, and Case Study
Original link: https://www.theverge.com/2017/8/30/16222850/youtube-google-brain-algorithm-video-recommendation-personalized-feed
Nice description of YouTube’s changed recommendation algorithm that was based on machine learning with some numbers. Feels like the issue remains that it works from past data: so once you watch a video about how to fix your washing machine, you may keep seeing videos of other washing machines.
Link to post
Posted on 2017 September 4 by polgarp · Tagged in Startups, Product management, and Overview
Original link: https://thepathforward.io/dos-and-donts-building-ai-startup/
Nice tips on how to build an AI based startup. Most of the advices would work for any type of technology based startups besides the ones on obtaining and working with data. So product people and designers your job will be very similar in an AI startup.
Link to post
Posted on 2017 September 4 by polgarp · Tagged in Poker, Infographics, and Overview
Original link: https://pokersites.me.uk/poker-ai/
Nice overview on how sophisticated AI systems fare against humans. Conclusion: don’t play against machines.
Link to post
Posted on 2017 September 4 by polgarp · Tagged in Bias, Image recognition, Ethics, and Machine learning
Original link: https://www.wired.com/story/machines-taught-by-photos-learn-a-sexist-view-of-women/
So sample datasets show more women in kitchen and men doing sports then vice versa. Machine learning algorithms learn it, and show us our existing biases. The more ML we use, the more we will see our biases in our data and everyday life - if we are carefully analyzing how ML systems behave. We realize the huge difference in the ideal world we are thinking about, and our real world, ML is data behaving. Here is the real question: how will this realization change our discussion around these biases? How will this change the discussions around biases in software development, in product development and design?
Link to post
Posted on 2017 September 4 by polgarp · Tagged in Use cases and Business model
Original link: https://hbr.org/2017/08/3-ways-companies-are-building-a-business-around-ai
Nice examples of how AI based companies work. Love the last one: “Package AI for the customer experience.” This is where the magic happen, features that work automagically and just make people’s life easier. Also the hardest - though have to wonder how the example company made use of focus groups.
Link to post
Posted on 2017 September 3 by polgarp · Tagged in Talk, Self-driving cars, and Ethics
Original link: https://www.ted.com/talks/iyad_rahwan_what_moral_decisions_should_driverless_cars_make
Interesting talk on the moral questions for self-driving cars. While cars are the most obvious examples on how algorithms can make decisions on who should die, these types of questions seem to appear in most domains.
Link to post
Posted on 2017 September 3 by polgarp · Tagged in Talk, Algorithmic discrimination, Weapons of math destruction, and Ethics
Original link: https://www.ted.com/talks/cathy_o_neil_the_era_of_blind_faith_in_big_data_must_end
Highly recommended talk by Cathy O’Neil on weapons of math destruction. It’s not just algorithms but any other part of products which can be seen as coded opinions - either unintentionally or intentionally by following a vision. Design seems well suited for helping data science, as it can clarify intentions, and make sure what we create actually matches our intentions.
Link to post
Posted on 2017 September 3 by polgarp · Tagged in Research, Social media, and Yelp
Original link: https://www.engadget.com/2017/09/01/ai-fake-yelp-reviews/
So after AI takes our jobs, it will also replace us as reviewers. Now the only thing we need is AI to be able to consume a plate of spaghetti.
Link to post
Posted on 2017 September 2 by polgarp · Tagged in Video, Overview, and AI future
Original link: https://www.youtube.com/watch?v=3Om9ssTm194
Nice explainer video on the potential dangers of super intelligent (general) AI. Good news: don’t be afraid of evil AI. Bad news: humans may be to AI as ants are to humans.
Link to post
Posted on 2017 September 2 by polgarp · Tagged in Product and Assistant
Original link: http://www.ahoy.ai/
Meeting scheduling is one of those tedious tasks that is well suited for automatization. Though there seems plenty of opportunities for using AI tech for more than just natural language processing - like how about finding meeting rooms, scheduling for ideal time to leave, not scheduling anything for lunch time etc.
Link to post
Posted on 2017 September 2 by polgarp · Tagged in Fun
Original link: https://www.theverge.com/tldr/2017/8/31/16234342/ai-so-hot-right-now-ysl-alexandre-robicquet
So wow. Just makes your average designer even more jealous of AI researchers.
Link to post
Posted on 2017 August 29 by polgarp · Tagged in Product, Design, and Graphic design
Original link: http://brandmark.io/logo-rank/
Although this tool gives very basic feedback, but is a nice example of how simple AI tools can enhance visual designers work. Just as doctors cannot review the huge body of existing medical research, designers cannot be aware of all the different logos in existence.
Link to post
Posted on 2017 August 28 by polgarp · Tagged in Music and Co-creation
Original link: https://www.fastcompany.com/40455600/this-new-ai-composed-pop-song-sounds-like-something-from-a-spotify-playlist
The title overacts a bit, while Taryn didn’t do composing herself, she did play with the parameters and set up the overall environment, like choosing tools somthing that can be arguably called composing. Similar to algorithmic art. I wonder when will this type of co-creation go more mainstream.
Link to post
Posted on 2017 August 28 by polgarp · Tagged in Google, Research, Analysis, and Drawing
Original link: http://research.googleblog.com/2017/08/exploring-and-visualizing-open-global.html
Amazing overview on the data available so far from the Quick, Draw! Google project. This type of data analysis should be more common, and a well understood tool of people for working with AI tech to combat bias. The cultural differences are also remarkable. While they are also a parameter machine learning can optimize for, but just stresses the need to collect data from all different groups of people.
Link to post
Posted on 2017 August 28 by polgarp · Tagged in Job automatization, Overview, and Research
Original link: https://medium.com/towards-data-science/artificial-intelligence-and-the-rise-of-economic-inequality-b9d81be58bec
A nice overview article on how the rise of AI (tech) will modify how people work, and what could be some of the impact on socieaty. It’s always fascinating to read such articles, as we are still at the very beginning of this transformation, and much of the prognostised changed can be influenced with systems designed for the needs of all involved.
Link to post
Posted on 2017 August 27 by polgarp · Tagged in Mozilla, Google, Open data, and Voice recognition
Original link: https://www.fastcompany.com/40449278/google-mozilla-and-the-race-to-make-voice-data-for-everyone
Pretty awesome that Mozilla now leading the way not just in free software, but also in free (and maybe open source) data. Software and algorithms are less and less interesting as a property, this is one of the reasons why Google and others release their internal tools. Software generally is also getting cheaper. But data is different, collecting a huge amount of data takes lots effort and resources and quality data (with low bias) even more so. This is why Mozilla’s efforst are so exiting, such open data (because how it’s crowdsourced) has a better chance of being low bias.
Link to post
Posted on 2017 August 27 by polgarp · Tagged in Weapons
Original link: https://futureoflife.org/autonomous-weapons-open-letter-2017
Short but important letter, on the topic of “let’s not build robots with guns”. Quite an impressive list of signatures. Terminator robots do sound scary, but hackable weapon systems even more so, especially if we look at the current state of IOT security. This is a cause even designers should push, let’s not design systems that can be used for weapons, and even lets not design systems that can be used to create weaponised AI tech.
Link to post
Posted on 2017 August 27 by polgarp · Tagged in Healthcare and Products
Original link: http://medicalfuturist.com/10-things-how-artificial-intelligence-could-make-me-a-better-doctor/
Lots of nice product ideas in this article by the Medical Futurist, with the simple premise of how AI tech could make a doctor better. Most of the ideas are quite simple (like better agenda scheduling) but in reality they are probably pretty complicated design tasks - not to mention there may be differences in how a doctor actually works compared to other knowledge workers. What makes this article super interesting, how simple compelling use cases can be formualted.
Link to post
Posted on 2017 August 20 by polgarp · Tagged in IBM, Watson, Healthcare, and Strategy
Original link: https://gizmodo.com/why-everyone-is-hating-on-watson-including-the-people-w-1797510888
Interesting longread on IBM’s Watson initiative. If I’d have to guess focusing marketing only on artificial intelligence (however companies call it) gets old really fast. As tech gets more mainstream companies and investors are less and less interested in it and mainstream has and maybe will have a hard time understanding it. It’s only a technology after all. So question is how fast can IBM turn it’s capabilities into truly astonishing products that could actually capture the attention of a wider audience.
Link to post
Posted on 2017 August 20 by polgarp · Tagged in Neural network, Fun, and Harry Potter
Original link: http://lewisandquark.tumblr.com/post/164291045392/harry-potter-and-the-word-level-recurrent-neural
Janelle always does fun stuff, this one with Harry Potter does not disappoint.
Link to post
Posted on 2017 August 20 by polgarp · Tagged in Design, Principles, and Data collection
Original link: https://uxdesign.cc/design-makes-ai-smarter-34a346e92b47
This post captures very well how important it is for systems with AI tech to have the right data, and also some of the design challenges that go with it. In the end the better design will result in more and reliable data.
Link to post
Posted on 2017 August 18 by polgarp · Tagged in Book
Original link: https://mitpress.mit.edu/books/what-algorithms-want
Seems like a super interesting book on designing with algorithms: “If we want to understand the gap between abstraction and messy reality, Finn argues, we need to build a model of “algorithmic reading” and scholarship that attends to process, spearheading a new experimental humanities.”
Link to post
Posted on 2017 August 18 by polgarp · Tagged in Product and Harassment
Original link: https://www.kiteai.com/
Nice looking service, lowering online harassment is good design goal. As the service argues it’s also good for UX. I wonder though why it needs to add “AI” to it’s name, especially since they don’t mention it anywhere on the main page.
Link to post
Posted on 2017 August 18 by polgarp · Tagged in OpenAI, Gaming, Dota 2, and Research
Original link: https://blog.openai.com/more-on-dota-2/
Some more details on how OpenAI trained their champion beating Dota 2 bot. Observing human behavior and playing against itself seems like nice strategies. For other applications the first one can be done pretty easily, I’m thinking a design bot reviewing iterations for example. But the second one can be more tricky - for design we would need to train the bot also to be able to do critiques.
Link to post
Posted on 2017 August 18 by polgarp · Tagged in Watermark, Google, Image manipulation, and Research
Original link: http://research.googleblog.com/2017/08/making-visible-watermarks-more-effective.html
Even with the solution offered in the article, it probably won’t take long to remove even the randomly warped watermarks. If anything this just shows the capability of AI tech to disrupt business models, in this case that of stockphoto sites and creators.
Link to post
Posted on 2017 August 14 by polgarp · Tagged in Manufacturing, Robots, and Job automatisation
Original link: https://www.fastcompany.com/3067414/robo-foremen-could-direct-human-and-robot-factory-workers-alike
So you decide what you want to build, provide a plan, and this tech can figure out from knowing the robot and human capabilities how to manufacture it. Maybe just one step ahead of telling another algorithm what you need, and it can figure out what’s the plan for it.
Link to post
Posted on 2017 August 14 by polgarp · Tagged in Philosophy, Ethics, AI apocalypse, and Research
Original link: https://thinkingwires.com/posts/2017-07-05-risks.html
Interesting article about the different dangers of future AI developments, and also some clearer descriptions what an AI apocalypse could be. This may be more of a philosophical question right now, but as we progress in improving AI tech, some safety should be probably built in.
Link to post
Posted on 2017 August 14 by polgarp · Tagged in TED talks, Tom Gruber, and Assistance
Original link: https://www.ted.com/talks/maurice_conti_the_incredible_inventions_of_intuitive_ai
Some generalities, but the segment of this talk on memory enhancement is pretty inspiring. It would be interesting to think about how human can deal with a total memory - remembering everything or most things, and how to design this experience to make it enjoyable to use.
Link to post
Posted on 2017 August 14 by polgarp · Tagged in OpenAI, Gaming, and Dota 2
Original link: https://blog.openai.com/dota-2/
Pretty big, and this might be a kindof Turing test too. Playing with somebody else, not communicating directly, only through the language of the game choreography.
Link to post
Posted on 2017 August 14 by polgarp · Tagged in DeepMind, Blizzard, StarCraft, Research, and Gaming
Original link: https://deepmind.com/blog/deepmind-and-blizzard-open-starcraft-ii-ai-research-environment/
Excellent resource to those who want to dabble in AI research while also looking for enjoyment from gaming. There are large datasets available, and you can also get your friends trying to beat you.
Link to post
Posted on 2017 August 13 by polgarp · Tagged in Google, Deep learning, and Browser
Original link: http://research.googleblog.com/2017/08/harness-power-of-machine-learning-in.html
Very cool, this library makes it super easy to play around with machine learning. Even to designers, since we already use js anyway for experimenting with frontend code. Also just as the article mentions, seems a nice way to quickly prototype interactions and products.
Link to post
Posted on 2017 August 13 by polgarp · Tagged in Policy, Overview, and Research
Original link: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3015350
Very through and interesting paper on the questions of policy making in relation to AI technologies. Even if AI related laws will be more affected by political discussions (like I expect a “Frankenstein Act” to happen as soon as a more complicated system does something dangerous to happen), many decisions will be based on research like this, so it’s well worth to understand what are some of the shaping ideas. Constraints set by laws will also affect the type of systems we will be able to design.
Link to post
Posted on 2017 August 13 by polgarp · Tagged in Assistant, Job automatization, Agents, and Productivity
Original link: https://blog.trello.com/artificial-intelligence-is-about-to-make-us-all-managers
“Bring Your Own Agent” is a nice description of something that is already happening. I tend to think about narrow AI systems as tools that make us smarter, just as traditional tools, even software make us stronger, more efficient and smarter. Only, AI systems make us much smarter. And just with earlier tools, you have to learn how to use your AI tools, and to customize to your own needs. Which makes the design of such tools partly the same, but also different, in a way that we have to design systems so people can understand and use easily the new complexities of AI tech.
Link to post
Posted on 2017 August 12 by polgarp · Tagged in Woman in tech, Assistant, and Bias
Original link: https://medium.freecodecamp.org/why-we-desperately-need-women-to-design-ai-72cb061051df
This article makes some nice points, and diverse teams do create better products. But there is actually more. Due to the nature of AI tech, it’s not just a strictly interface question (like pre-programmed answers in an assistant) if a product can be well designed without women. The data collection and learning methods are just as important and can make or break a diverse enough product. So if anything issues from not having women (or diverse teams in general) in product teams will just grow in magnitude.
Link to post
Posted on 2017 August 12 by polgarp · Tagged in Co-design, Interaction design, and Product design
Original link: https://medium.com/swlh/invisible-design-co-designing-with-machines-aea62a1e0f6d
“Invisible Design — a process and design language for product designers working with artificial intelligence and technologies like machine learning.” I like this article, as it stresses the necessity of collaboration between functions (for example data and design) for successful AI products. But this is also true for almost all imaginable products, good experiences are not born in isolation.
Link to post
Posted on 2017 August 12 by polgarp · Tagged in Course, Deep Learning, Andrew Ng, and Coursera
Original link: https://www.coursera.org/specializations/deep-learning
A new series of courses and specialization on Coursera headed by Andrew Ng. His Machine Learning course is one of the best introductions - very much understandable for less technical persons (you do need to know a little bit about matrix multiplication though).
Link to post
Posted on 2017 August 11 by polgarp · Tagged in Agriculture, Image recognition, and DIY
Original link: http://www.newyorker.com/tech/elements/diy-artificial-intelligence-comes-to-a-japanese-family-farm?utm_source=nextdraft&utm_medium=email
I think the home DIY application of AI tech is still some time away, Koike after all was an engineer. But with the mobiles soon coming out with dedicated ML chips, this could all change. Maybe just like 3D printing, “intelligent” components will be soon of-the-self anybody can hack together. I wonder what perspectives does this open for people.
Link to post
Posted on 2017 August 11 by polgarp · Tagged in Healthcare and Machine learning
Original link: https://www.wired.com/story/ai-that-will-crowdsource-your-next-diagnosis/
This is a nice example how a system built with AI tech can gradually learn from human participants and get better over time, provide services that would be otherwise not possible (these patients couldn’t afford the diagnosis they would have needed) and still not act as “job stealer” - the specialist just have more time to look after patients really in need.
Link to post
Posted on 2017 August 7 by polgarp · Tagged in Drawing, Machine learning, Co-creation, and Project
Original link: http://nono.ma/suggestive-drawing
A fascinating project on how humans and AI powered bots can co-create together. There are lots of interesting details on how the ML algorithms were trained, and how working together affected the creator’s art style.
Link to post
Posted on 2017 August 7 by polgarp · Tagged in Research, TED talk, and Computer vision
Original link: https://www.ted.com/talks/joseph_redmon_how_a_computer_learns_to_recognize_objects_instantly/
Pretty amazing results on real time object detection, also a fun to watch TED talk. Since Darknet/Yolo is free to use and it seems to work reliably on phones, this will be especially useful for applications in the AR space.
Link to post
Posted on 2017 August 7 by polgarp · Tagged in Design, Product, Persona, and Interaction style
Original link: http://uniform.net/blog/june-2016/designing-with-ai/
One of the challenges of designing products with AI tech is how the interaction between the humans and the machine should work. We have pretty solid understanding on this works with “dumb” software, for optimal usability we want the user to be in control. But with smart algorithms, the user being in control may not lead to the right outcomes. I like the personas, or rather models of the interaction described in this article, they seem pretty useful when defining what should be the basic approach of a system.
Link to post
Posted on 2017 August 6 by polgarp · Tagged in Product management, Startups, and Getting started
Original link: https://chatbotsmagazine.com/product-management-for-ai-startups-d738aebb8430
A little bit chaotic, but this is kinda ok overview on how to get started with AI products. There are plenty of things missing though - like for example how to get feedback on your implementation (ok this one is easy, just hire some UX people), or how to align your team and your stakeholders around a AI vision (seems harder than on traditional products), or how to know when your AI tech based product/feature just failed (pretty hard, since there can be many variables on learning speed, data hygiene etc).
Link to post
Posted on 2017 August 6 by polgarp · Tagged in Business and Report
Original link: http://www.mckinsey.com/business-functions/mckinsey-analytics/our-insights/how-artificial-intelligence-can-deliver-real-value-to-companies
The included report available to download is more interesting. So early adopter companies push hard into AI tech, while those who don’t risk to be left back. There is definitely a place for most industries to improve their user experience, not just on the “traditional” user facing channels (I guess a huge opportunity for enterprise focused startups).
Link to post
Posted on 2017 August 6 by polgarp · Tagged in Facebook, Machine learning, and Translations
Original link: https://www.theverge.com/2017/8/4/16093872/facebook-ai-translations-artificial-intelligence
Quite impressive, given Facebook’s scale, and ofcourse they have access to a large dataset of human written language: posts, comments and chats. So this gives the chance to any two people to better understand each other, across languages and even subcultures. Some pretty neat possibilities to make communication between people easier.
Link to post
Posted on 2017 August 6 by polgarp · Tagged in Policy, Research, and Politics
Original link: https://www.theverge.com/2017/8/3/16007736/china-us-ai-artificial-intelligence
Interesting to compare the current research and development in AI tech to the space race and there are actually quite a lot of parallels.
Link to post
Posted on 2017 August 1 by polgarp · Tagged in Machine Learning, Attacks, and Research
Original link: https://arxiv.org/pdf/1707.08945v1.pdf
It’s interesting to see more and more research published on how to fool machine learning systems, vision based in particular. I can see designers on both sides, on one had as more and more ML tech is deployed to follow people without their informed consent there is a need for systems enabling users to hide. On the other hand systems designed should protect well-intentioned users from harm coming from attackers.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in Design, General AI, Narrow AI, and Introduction
Original link: https://uxdesign.cc/you-can-be-an-ai-designer-46a0fd45f47d
As far as different categories of AI tech from narrow to superintelligence, there are difference skills we will need to actually design these products and systems. Even AGI feels like sci-fi at the moment, but all the experience in design we will gather from working with narrow intelligence products should serve well.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in Research, Co-creation, Machine learning, and Training data
Original link: https://arxiv.org/abs/1707.05173
A critical questions of AI systems is how to stop them to make potentially catastrophic decisions. This danger comes from such systems optimizing for reaching a specific goal, without minding any other casualties on the way there. This paper offers a way to solve this problem, by introducing human oversight into the training. This is an other example to show how effective is for humans working together with AI tech, rather then the tech substituting people.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in Overview and Introduction
Original link: https://hbr.org/cover-story/2017/07/the-business-of-artificial-intelligence
Cover story on HBR, it’s more of an other general introduction to the current batch of AI tech, but there are actually a few tasty details.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in AI-First, Business model, and Organization
Original link: https://medium.com/@jeremy_p_barnes/the-ai-first-business-model-fcc41c069440
There seems to be a lot of talk about “AI-first business”, though much less info on how exactly this works. This article offers some points on this, in particular how such companies should function. Unfortunately this model misses the design function, though in most companies it’s probably rolled into the product organization. This setup would lower the value of people-facing AI products, as the design knowledge would be missing from the C-Level.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in Mozilla, Voice recognition, and Open data
Original link: https://www.theverge.com/2017/7/27/16019222/open-source-data-ai-mozilla-common-voice-project
Using products built on open data is probably more important than using products built on open software, since as independent review makes it more apparent if a company works with dark patterns the same is true for working with biased data. This also enables to design products without a large supporting data collection mechanism.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in Design, Interaction design, and Principles
Original link: http://interactions.acm.org/archive/view/july-august-2017/intelligence-on-tap#new_tab
AI tech as a new type of material to design with is an interesting idea: just as with other types like software, you have to understand how things work. But even to software, the workings are different, and might be even counter-intuitive to experienced designers. The challenges listed in this article also show that there is plenty of work left in defining new types of interactions when working with “AI material”.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in Privacy, Attacks, and Fashion
Original link: https://www.rbth.com/science_and_tech/2017/07/22/hiding-from-artificial-intelligence-in-the-age-of-total-surveillance_808692
Interesting read, not just because of the Russian angle. Hiding from facial recognition is hard online, and getting harder in real life too. But also might be more and more interesting to more and more people, as keeping privacy, and staying anonymous for ad targeting and “experience personalization” will get more important. I wonder if this will have a real fashion impact in the future.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in Design and Machine learning
Original link: https://hackernoon.com/creating-brand-personas-with-machine-learning-647d9314baaa
A neat project describing a concrete instance how machine learning can be used to generate original research. Most design research will remain person-to-person and qualitative in the future too, but applications like this can enhance findings.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in Policy and Chine
Original link: https://www.wired.com/beyond-the-beyond/2017/07/chinese-state-council-guidelines-artificial-intelligence/
Pretty interesting. While even companies struggle to get a unified vision on how they’ll apply AI teach, Chine elevates this to a state level. I’m not sure how effective is this to drive research forward, but they do add enough resources to make it happen.
Link to post
Posted on 2017 July 30 by polgarp · Tagged in DeepMind, Research, Agents, and Imagination
Original link: https://deepmind.com/blog/agents-imagine-and-plan/
Posted on 2017 July 26 by polgarp · Tagged in Product management, Product development, and Startup
Original link: https://medium.com/@matyas_82059/the-exciting-challenges-of-developing-an-ai-product-e04cce4cba58
There are too few articles about the challenges of developing actual products with AI tech like this one.
Link to post
Posted on 2017 July 25 by polgarp · Tagged in Overview, General AI, and AI Doomsday
Original link: http://togelius.blogspot.hu/2017/07/some-advice-for-journalists-writing.html
Not just for journalists, really for anybody getting into this field. Since there is a lot of discussion going on what a general AI could be, most people loose sight on what is actually realistic right now.
Link to post
Posted on 2017 July 23 by polgarp · Tagged in Video game, Design process, and Simulation
Original link: http://www.karineskoog.com/thehunter-call-of-the-wild-designing-believable-simulated-animal-ai/
A fascinating article about how to create realistic simulated behavior. While this all seems very specific, it’s kinda the same amount of research that goes into any product with similar complexity. Just as with any design process, there should be a significant effort spend to understand the domain and subject matter, to have enough data to build a useful (or in this case, fun) level of algorithmic intelligence.
Link to post
Posted on 2017 July 23 by polgarp · Tagged in Ethics, Bias, and Policy
Original link: https://joanna-bryson.blogspot.hu/2017/07/three-very-different-sources-of-bias-in.html
Many of the ethics issues present when working with AI tech is similar to the issues designers face when we create products. Although the design approach is implicitly humanistic, we still haven’t figured out sure ways to follow specific ethical guidelines - something that doesn’t bode well for products with AI tech. On the other hand ethical approaches are integrated into our methods, like design research - which enables us to have a headstart when working with AI tech.
Link to post
Posted on 2017 July 23 by polgarp · Tagged in Research, Cooking, and Computer vision
Original link: https://www.theverge.com/tldr/2017/7/20/16005826/mit-csail-recipes-ai-neural-network-algorithm
Nice project - with so many recipe sites with pictures on the internet there should be enough data to do this just right. Would be a useful product, but maybe less so than for example Shazam.
Link to post
Posted on 2017 July 23 by polgarp · Tagged in Art, Design process, Case study, Creativity, and Co-creation
Original link: https://artplusmarketing.com/digital-processes-inspiring-analog-paintings-a358eb7801a0
Loved this description of a creative process. There are more and more examples emerging where a humans and algorithms support each others in creative processes. Wonder how effective this type of collaboration would be for a designed experience.
Link to post
Posted on 2017 July 23 by polgarp · Tagged in Education, Product, and Overview
Original link: https://thenextweb.com/artificial-intelligence/2017/03/13/how-artificial-intelligence-enhances-education/
Nice overview of AI applications in education. This is one of the fields that could benefit form AI tremendously, as it would would need huge amount of knowledge transfer (mostly methods between organisations) and human effort (more personalized teaching needs more teachers), while is low on capital, so automatization would be beneficial. However the long term impact on the students would be huge, so careful design is called for.
Link to post
Posted on 2017 July 23 by polgarp · Tagged in Elon Musk, Policy, and AI Doomsday
Original link: https://www.theverge.com/2017/7/17/15980954/elon-musk-ai-regulation-existential-threat
If strict regulation would happen (armed Federal AI Bureau officers cracking down on rouge researchers), not sure if it would amount to anything. AI research has an even lower cost than other discipline with a high potential for destruction (like biotech for example). Especially with the examples mentioned, like propaganda. So we need to design systems with thinking about malicious entities who would use the advances against us. Asimov’s laws of robotics feels more and more attractive.
Link to post
Posted on 2017 July 20 by polgarp · Tagged in Product and Analytics
Original link: https://www.producthunt.com/posts/paveai-2-0
“Turn Google Analytics into actionable insights using A.I.” Understanding product usage data, or really any type of analytics suits automatization well. You need to apply tons of heuristics, recognize patterns, trends, make predictions. Seems not that hard to design systems that helps non-professionals act on their data. Plenty of research needed though on both sides too - to understand the heuristics from professionals and to understand needs of the end users.
Link to post
Posted on 2017 July 19 by polgarp · Tagged in Adversial, Research, and OpenAI
Original link: https://blog.openai.com/robust-adversarial-inputs/
Fooling AI tech will be the hacking of the future - and it may be also part of individuals increased privacy. Such research will be more common as an algorithmic arms race will ensue, similarly to today’s ad blocking apps.
Link to post
Posted on 2017 July 19 by polgarp · Tagged in Checkers, Game, Humans vs computers, and History
Original link: https://www.theatlantic.com/technology/archive/2017/07/marion-tinsley-checkers/534111/
Now more of a piece of history how checkers was won by computers, still a fascinating story about the people on both sides.
Link to post
Posted on 2017 July 19 by polgarp · Tagged in Apple, Machine Learning, and Research
Original link: https://machinelearning.apple.com/
Probably most of the contents will be only interesting for ML researchers, still interesting to keep an eye out what Apple is working on. Also shows their commitment to this tech, and a bit of a change to Apple’s general approach of not really publishing any research results.
Link to post
Posted on 2017 July 15 by polgarp · Tagged in Chatbot, Assistant, and Legal
Original link: https://www.theverge.com/2017/7/12/15960080/chatbot-ai-legal-donotpay-us-uk
This looks like a very useful service. “Lawyerspeak” can differ from plain, everyday language by a lot, since it’s very precise and specific. This specificity also makes it ripe for automation, and lawyer firms are also doing it via templates especially in trivial cases. But what is even more awesome that this automation helps a wider range of people not able to afford professional help.
Link to post
Posted on 2017 July 15 by polgarp · Tagged in Machine learning, Bias, and Discrimination
Original link: https://www.recode.net/2017/7/13/15967456/women-tech-business-sexual-harassment-discrimination-trump-catriona-perry
I’m not sure if women’s are better off in the design community than the tech - especially if they work as designers in a tech company. While it seems one of the most important factor of designing for AI tech is how we handle data bias, just following guidelines is probably not enough. Organisation are only capable to produce software that mirrors them, so design teams have to consist of equal ratio of men and women.
Link to post
Posted on 2017 July 15 by polgarp · Tagged in Google, Photography, and Co-create
Original link: https://research.googleblog.com/2017/07/using-deep-learning-to-create.html
This would be another result, where AI tech helps non-professionals achieve professional results. In photography in particular this feels as big change as the appearance of digital photography, it brings a new way of creating photos to people. So how will this affect how we remember things?
Link to post
Posted on 2017 July 15 by polgarp · Tagged in Google, Research, and Interaction design
Original link: https://www.blog.google/topics/machine-learning/pair-people-ai-research-initiative/
A new program by Google, “focusing on the human side of AI”, so essentially researching how interaction happens between people and AI tech. This sounds more like an academic program (I expect nice papers on next year’s CHI), but they seem to take a route that will be also useful for a wider range of practitioners, promising to release guidelines and open source tools. Probably something worth keeping an eye on.
Link to post
Posted on 2017 July 15 by polgarp · Tagged in Google, Principles, Methods, and Interaction design
Original link: https://medium.com/google-design/human-centered-machine-learning-a770d10562cd
While the title feels a bit philosophical, this article from googlers contains some quite practical advice and even a few concrete methods for UXers on how to design machine learning products. It’s also cool how Google internally trains UXers on how to work with these technologies.
Link to post
Posted on 2017 July 15 by polgarp · Tagged in Product management, AI-first, and Product development
Original link: http://www.mindtheproduct.com/2017/07/building-products-ai-first-aparna-chennapragada/
Interesting perspective, however I like the idea of having to train the product instead of merely build. If anything this just means there needs to be more understanding on the users (or even more generally the people) to create successful products, so yay for UX people.
Link to post
Posted on 2017 July 15 by polgarp · Tagged in Experiment and Product
Original link: https://www.theverge.com/tldr/2017/7/10/15946296/amazon-bot-smartphone-cases
Even if this is not case, it could be. Just shows how common, basic graphic design will become mostly the domain of AI users, rather then design operators.
Link to post
Posted on 2017 July 14 by polgarp · Tagged in Chess, Go, Board games, and History
Original link: https://fivethirtyeight.com/features/the-bots-beat-us-now-what/
Posted on 2017 July 12 by polgarp · Tagged in Microsoft, Computer vision, and Disability
Original link: https://www.microsoft.com/en-us/seeing-ai/
Pretty amazing app from Microsoft, it “narrates the world around you. Designed for the low vision community, this research project harnesses the power of AI to describe people, text and objects.” Just as with other types of personal assistants, AI tech may take over helping people with disabilities. Great way to enhance people’s skills.
Link to post
Posted on 2017 July 11 by polgarp · Tagged in Chatbots and Principles
Original link: https://latenightcoding.co/chatbots-customer-service/
This article hits the right pain points with chat bots on many levels. Pre scripted bots are vastly different to real assistants like Siri, just way more limited. But the costs are also much lower. They seem to be still useful for specific contexts, so you need to research and design properly, just like with any other channel.
Link to post
Posted on 2017 July 9 by polgarp · Tagged in Gambling, Sport, and AI-first
Original link: https://www.theverge.com/2017/7/6/15923784/ai-predict-sport-betting-gambling-stratagem
I’d call the business practices controversial too, but looking behind their revenue model it’s interesting to see how they transform their operation from human-run to mostly AI run one step at a time. This could be a good pattern to follow for other companies that rely heavily on human analysis and decision making.
Link to post
Posted on 2017 July 9 by polgarp · Tagged in Job automatization, Co-create, and Self driving
Original link: https://www.bloomberg.com/news/features/2017-06-22/these-truckers-work-alongside-the-coders-trying-to-eliminate-their-jobs
A nice longread about a job we are trying automatize, excellent to raise empathy and to start thinking about how individual new products will transform industries, and more importantly transform people’s lives. If anything, this emphasizes the need for universal design in the case of AI products: service, digital and organizational design should go hand in hand.
Link to post
Posted on 2017 July 9 by polgarp · Tagged in Overview, Art, Creative, and Co-create
Original link: https://attentionecono.me/creative-artificial-intelligence-music-writing-video-movies-2017-7e9c192594a3
A nice overview on the application of AI tech in the creative fields (part 1). It’s fascinating to see the different approaches companies and projects take, in terms of how they use AI tech. AI-human collaboration seems to be the most successful.
Link to post
Posted on 2017 July 9 by polgarp · Tagged in App and Photography
Original link: https://9to5mac.com/2017/07/07/prisma-stick-ai-iphone-ipad/
I liked Allo’s solution of actually generating a more cartoon looking avatar.
Link to post
Posted on 2017 July 9 by polgarp · Tagged in Website and Blog
Original link: https://inspiring.ai/
Nice source of news of AI + data + design.
Link to post
Posted on 2017 July 9 by polgarp · Tagged in Google, Journalism, and Automatization
Original link: https://www.recode.net/2017/7/7/15937436/google-news-media-robots-automate-writing-local-news-stories
An interesting project, and surely a good way to move forward for media outlest. Of course the question is how well these stories will be liked by consumers.
Link to post
Posted on 2017 July 9 by polgarp · Tagged in Instagram and Fashion
Original link: https://www.technologyreview.com/s/608116/data-mining-100-million-instagram-photos-reveals-global-clothing-patterns/
Instagram has been long the treasure trove for a design researcher, but such scale is super exciting. I wonder what other type of research is becoming possible with this.
Link to post
Posted on 2017 July 9 by polgarp · Tagged in Research, Psychology, Deep neural networks, and Bias
Original link: https://arxiv.org/abs/1706.08606
An awesome research proposing to analyse AI systems with psychological methods. While understanding such systems seems to be a difficult problem now, we have lots of knowledge about how human minds work. And while the two might be not function similarly at all, psychology does provide a nice toolset, and much of design research is applied psychology, since it works with humans too, One more place where existing design (research) knowledge might be useful.
Link to post
Posted on 2017 July 8 by polgarp · Tagged in Enterprise, B2B, and AI-first
Original link: https://techcrunch.com/2017/06/28/the-next-generational-shift-in-enterprise-infrastructure-has-arrived/
Interesting analysis on how enterprise software might incorporate AI tech - essentially into a new layer between data and functions called referred to as System of Intelligence. Conceptually this seems the right place - not a piece of UI labelled as AI, but more like a backend kindof that works together with the human facing layers. The analysis provides a few nice examples too, for example how these kind of systems might change security processes. I’m not so sure about the name (SoI), maybe this one will be also corporate bullshit very soon.
Link to post
Posted on 2017 July 2 by polgarp · Tagged in Overview and Business model
Original link: https://machinelearnings.co/winning-strategies-for-applied-ai-companies-f02cac0a6ad8
Nice overview on the potential of AI companies.
Link to post
Posted on 2017 July 2 by polgarp · Tagged in City planning
Original link: http://meetingoftheminds.org/smart-cities-artificial-intelligence-balancing-opportunity-risk-21706
Cities are a great achievement of humanity, so it stand to reason AI tech can truly augment how we build and live in our cities. Plenty of design challenges.
Link to post
Posted on 2017 July 2 by polgarp · Tagged in Prediction and Recommendation
Original link: http://reallifemag.com/instant-recall/
Whenever you get a product recommendation or some kind of prediction through AI tech, it’s always just an extrapolation of your past self, not based your actual state. That is why ads from a product category follow you around the web, even if you already bought the said product and don’t wanna think about it any more. This is a though issue to solve, as you need to really understand humans and how they function to be able to design such systems.
Link to post
Posted on 2017 July 1 by polgarp · Tagged in Deep learning and Overview
Original link: http://rbharath.github.io/what-cant-deep-learning-do/
I think it’s important for designers to understand the limitations of AI tech (just as we would understand the limitations of any type of tech / org we are working with). This article collects nicely what are some of the challenges with current deep learning systems (one of the more popular / powerful type of machine learning today).
Link to post
Posted on 2017 July 1 by polgarp · Tagged in AI-first, Business, Business model, and Product management
Original link: https://medium.com/intuitionmachine/monetizing-artificial-intelligence-d69fe9175785
Although a bit speculative, this article describes a nice way to build AI products (artificial-artificial intelligence): find a problem that requires basic cognition that is currently done by humans and somebody is paying for it, start automatizing parts, apply the learned experience for similar problems. Just as if the customer would be talking to a human, the initial conversation can be designed to be at least as usable or maybe even more usable than plain communication.
Link to post
Posted on 2017 June 30 by polgarp · Tagged in Ethics and Principles
Original link: https://www.fastcompany.com/40436189/5-principles-to-make-sure-businesses-design-responsible-ai
I like these principles, they can be even viewed as a more specific framework for product design of such applications. It’s also great, that all of these principles focus on the connection of AI tech and humans, just as the high level ideas of HCI.
Link to post
Posted on 2017 June 29 by polgarp · Tagged in Instagram and Social media
Original link: https://www.wired.com/story/instagram-launches-ai-system-to-blast-nasty-comments/
So Instagram will be a nice place to use, only the other 99% of the Internet remains. As such algorithms will get more widespread, I wonder how it will change the course of human discussions. Internet definitely changed this, people are more aggressive in online discussions, party since the emotional value of written text is different. Can such an algorithm overtake what is being said?
Link to post
Posted on 2017 June 29 by polgarp · Tagged in IBM, AI doosmday, and Policy
Original link: https://www.recode.net/2017/6/27/15875432/ibm-congress-ai-job-loss-overlord
I wonder what story IBM’s people tell about what kind of changes are to be expected. Both trying to stop organizations expand their AI tech’s use, or trying to fit AI tech into current policy constraints seems to be a mistake, it can have a much more profound effect on people and society as a whole. Lawmakers shouldn’t be ignorant even on the details of this tech.
Link to post
Posted on 2017 June 25 by polgarp · Tagged in Image generation
Original link: http://alteredqualia.com/xg/examples/eyes_gaze3.html
Freaky demo of machine generated faces following the cursor movement. Feels almost at the border of the uncanny valley, still somehow not looking human enough.
Link to post
Posted on 2017 June 25 by polgarp · Tagged in Andrew Ng
Original link: https://www.deeplearning.ai/
A new project by Andrew Ng. No info yet, but should be interesting. Coming in August.
Link to post
Posted on 2017 June 25 by polgarp · Tagged in AI development
Original link: https://www.fastcompany.com/40433959/why-the-military-and-corporate-america-want-to-make-ai-explain-itself
Of course you wanna now why something happened if that thing ended up as a mistake, not even to know whom to blame, but to not make the same mistake twice. Especially if the mistake costs you. There are two issues. One, even if the tech would be capable to explain itself, it’s not clear that anybody besides the smartest AI researchers would be able to understand it. Second, all this just kinda assumes that people (as opposed to a piece of AI tech) are capable of explaining their actions all the time.
Link to post
Posted on 2017 June 25 by polgarp · Tagged in Chatbots and Overview
Original link: https://www.recode.net/2017/6/22/15855492/ai-artificial-intelligence-nonprofit-good-human-chatbots-machine-learning
Some really nice examples of AI tech deployed by nonprofits to help people. Besides the point that such applications are now possible even without massive expenses and profit goals by corporations (both data and tech is available and affordable), AI tech can help many such organizations to scale up, do more with the same amount of resources. Feels like many design challenges remain though, it’s not easy to automatize care.
Link to post
Posted on 2017 June 25 by polgarp · Tagged in AI policy
Original link: https://80000hours.org/articles/ai-policy-guide/
Interesting article on how to start a career in AI policy, and especially what questions are there to work with. These questions should be also worth thinking about when you work in a more defined context, for example on product or in a company working with AI tech. So for product managers (business questions) and designers (user questions).
Link to post
Posted on 2017 June 25 by polgarp · Tagged in Fun
Original link: http://lewisandquark.tumblr.com/post/162097037117/ancient-wisdom-from-the-neural-network
These generated wisdom proverbs are pretty awesome, most of it could actually end up on the Facebook wall, like “No sweet is half the barn door after the cat.”
Link to post
Posted on 2017 June 20 by polgarp · Tagged in AI design and Design patterns
Original link: http://www.uxmatters.com/mt/archives/2017/06/predictive-user-experience.php
A deep article describing the use of predictive modelling in design, and also provides some patterns around the anticipatory design pattern. Since this is a frequent use of AI tech, it’s definitely worth reading.
Link to post
Posted on 2017 June 18 by polgarp · Tagged in AI adoption
Original link: https://hbr.org/2017/06/if-your-company-isnt-good-at-analytics-its-not-ready-for-ai
While this article talks about companies adopting AI tech into their internal services, I feel pretty much the same is going on when building a product with AI tech, or more importantly if a product’s core capability is AI. You have to design data collection and basic automation too if you want the product to be successful.
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Startups, Business model, and Enterprise
Original link: http://www.bradfordcross.com/blog/2017/6/13/vertical-ai-startups-solving-industry-specific-problems-by-combining-ai-and-subject-matter-expertise
If you are looking to create an AI (or really any technological) startup, this will be an informative article + talk. It gives a broad framework on how to choose opportunities, target market and industry, especially if you are looking for the enterprise market.
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Google and Energy
Original link: https://blog.google/products/earth/using-machine-learning-help-people-make-smart-decisions-about-solar-energy/
This is a nice story for two reasons. First it shows how Google makes a positive impact on it’s surroundings. Second it shows how even application’s that need a lot of content - like a database with all the solar installations in a region is much easier to create with machine learning. You don’t necessarily need as much resources or have a well designed crowd source product.
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Music
Original link: https://www.theverge.com/2017/6/15/15807056/shimon-ai-robot-deep-learning-marimba
I guess as with similar efforts the lack of feedback doesn’t help to steer the results into a direction - as opposed to for example with Go or similar games.
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Military, Robots, and Overview
Original link: https://www.fastcompany.com/3069048/where-are-military-robots-headed
An overview of current (public) military use of AI tech - but mostly on tactical autonomous vehicles.
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Google and Journalism
Original link: https://www.recode.net/2017/6/13/15789178/new-york-times-expanding-comments-artificial-intelligence-google
Many interesting details how NYT’s practices may collaborate with AI tech, and even more interesting opportunities. Suppressing hate speech is a worthy design challenge in itself, but also surfacing further information based on user generated content that can form the basis of other stories feels like a great use case. So it’s not just a bunch of learning systems that need to work well, but also all the interfaces for the human collaborators need to be done well to stay effective on scale.
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Microsoft and Video games
Original link: https://www.theverge.com/2017/6/14/15801700/microsoft-ai-perfect-ms-pac-man-score
“designated a top agent (Microsoft likens this to a senior manager at a company)” - ofcourse MS likened this to a senior manager. Two interesting things here: First, can human organization best practices be applied to how machine learning systems are organized? Second, could we find better organizational best practices through watching AI systems collaborating?
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Customer management and Overview
Original link: https://thenextweb.com/contributors/2017/06/12/artificial-intelligence-revolutionizing-customer-management/#.tnw_dObLu3To
A few examples of companies using AI tech in CRM services. Even if companies keep their human agents talking to customers, there are many way AI systems can contribute to a more pleasant experience. An other way of thinking about these, what would you rather have, a “robot” solving your issue fast, or talking to a much slower human operator?
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Design tool and Fonts
Original link: http://fontjoy.com/
Nice little tool for selecting fonts. While selecting matching fonts is one of the more fun activities of typography, it’s also though to get it right and I’m guessing many times designers are happy to go through this part of the process faster.
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Facebook and Chatbots
Original link: https://www.recode.net/2017/6/14/15796812/facebook-bots-ai-artificial-intelligence-research
Kind of expected that AI systems will be better than humans at negotiating, as they are able to keep more parameters in mind, and can get clearer focus. Looks like instead we are teaching them how they can be more like us.
Link to post
Posted on 2017 June 17 by polgarp · Tagged in Voice UI and Interaction
Original link: https://medium.com/@MrAlanCooper/alexa-please-kill-me-now-eb693ce73258
While conversational user interfaces are not an AI technology, they have became prevalent in recent days due to voice recognition and NLP allowing it’s wider use. Alan Cooper makes a strong case on how these UIs still fail to capture user intention, mostly since human’s communication is not clear in itself.
Link to post
Posted on 2017 June 16 by polgarp · Tagged in Kickstarter and Product
Original link: https://www.kickstarter.com/projects/910224259/xtnd-worlds-first-electric-board-with-ai
Seems like “physical product with AI in it” is the new “physical product connected to the internet”. Even if this is only Kickstarter, it may show the future where things with embedded / ambient AI actually make our life easier - as opposed with internet connected things only collecting data.
Link to post
Posted on 2017 June 16 by polgarp · Tagged in Chatbot and Web design
Original link: https://firedrop.ai/
“Forget drag and drop, just talk to Sacha”. Interesting product, conversion with bot trying to replace more traditional interactions. Also a nice example how design operational tasks, like simple web design are automatized.
Link to post
Posted on 2017 June 14 by polgarp · Tagged in Product, Trello, Assistant, and Productivity
Original link: https://butlerfortrello.com/
Personal productivity is ripe for innovation with AI tech. This Trello automation tool seems promising - while most features are just a some cool automation that the user can define the best feature seems to be the smart generation of automation options. Most often it’s not obvious what actions could be done faster to most people, so this one seems hugely useful. (I use Trello all the time).
Link to post
Posted on 2017 June 13 by polgarp · Tagged in Emotion
Original link: https://www.technologyreview.com/s/608036/we-need-to-talk-about-the-power-of-ai-to-manipulate-humans/
Sounds just about right that people are more likely to form relationships with AI systems designed to form relationships than with an other human being. This just stresses that product with AI tech has to look for a positive change in human behavior, that is positive not (only) for the product’s KPIs.
Link to post
Posted on 2017 June 13 by polgarp · Tagged in Chatbot, Healthcare, and Mental well-being
Original link: https://www.wired.com/2017/06/facebook-messenger-woebot-chatbot-therapist/
Interesting product with some potential. It kinda highlights that AI can help in completely new fields, where “scaling up” (providing certain social services to more people) is just simply not viable.
Link to post
Posted on 2017 June 13 by polgarp · Tagged in Infographics and Introduction
Original link: http://blog.solvvy.com/blog/the-basics-of-artificial-intelligence-decoded
Nice but very basic infographical overview on AI basics.
Link to post
Posted on 2017 June 13 by polgarp · Tagged in Future AI
Original link: https://hbr.org/2017/06/how-to-prepare-the-next-generation-for-jobs-in-the-ai-economy
Interesting to do some light predictions on how new technologies will affect education on the long term. These tips from HBR however sound just like the ideas from earlier: learn problem solving and coding (to understand computers). I think we can expect even bigger impact, for example learning about AI as we do currently about biology, a basic topic that children need to know about to understand even how their immediate context works.
Link to post
Posted on 2017 June 13 by polgarp · Tagged in Aviation
Original link: https://www.wired.com/story/boeing-autonomous-plane-autopilot
A fully automated plane is not as terrifying, as an automated plane with mandatory human oversight. This would mean once again systems, where the easiest, routine tasks are automated, and humans are expected to handle to more tricky situations - when they are bored, out of routine, without any opportunity to learn by experience, and without any warm up. Leads to the classic HCI situation, where tons of usability work needed to avoid errors.
Link to post
Posted on 2017 June 13 by polgarp · Tagged in AI design and Principles
Original link: https://bigmedium.com/speaking/design-in-the-era-of-the-algorithm.html
Josh Clark’s transcript of a talk he did on algorithmic design and “algocracy”. The best part is not even the 10 pretty great design principles he proposes for data-driven products or the tons of awesome examples, but how he describes the potential for both designers and researchers to have a positive impact on these new type of products.
Link to post
Posted on 2017 June 11 by polgarp · Tagged in Apple
Original link: https://www.technologyreview.com/s/608051/tim-cook-technology-should-serve-humanity-not-the-other-way-around/
Probably correlating to WWDC, Apple seems to shed a little bit of light on the AI work they’ve been doing. Not much new info in this piece, just a few “FYI Apple is in the AI game”s.
Link to post
Posted on 2017 June 11 by polgarp · Tagged in Food production
Original link: https://www.fastcompany.com/40419891/these-food-computers-use-ai-to-make-climate-recipes-for-the-best-tasting-crops
The idea of having total control over a crop’s environment is kind of every farmer’s dream. But the more amazing idea here is to also being able to share these recipes with others, so growing crops gets to be even less of a guesswork (just based on common wisdom).
Link to post
Posted on 2017 June 11 by polgarp · Tagged in Apple
Original link: http://deepdojo.com/apple-introduces-core-ml
Maybe a technical detail on Apple’s machine learning tech, but still useful to know. If you want to include AI tech in your product (mobile app), iOS maybe the easier way to deploy first.
Link to post
Posted on 2017 June 10 by polgarp · Tagged in Apple
Original link: https://backchannel.com/an-exclusive-look-at-how-ai-and-machine-learning-work-at-apple-8dbfb131932b
Fascinating read on Apple’s AI capabilities. There are plenty of interesting details, like how the machine learning technology is integrated into products, and into product teams, how Siri is at the center of most of these efforts. And most importantly, how Apple works with the users’ data, while keeping it’s privacy ethos.
Link to post
Posted on 2017 June 10 by polgarp · Tagged in Ethics and Startups
Original link: http://tomtunguz.com/ethics-in-startups/
It’s the designers job to make sure the products they create are ethical, even when using AI tech - but such obligations are rather inspirational than practical. So I love the idea to view ethics more pragmatically and start creating products that protect users from evil software, instead of just expecting software creators to behave.
Link to post
Posted on 2017 June 10 by polgarp · Tagged in Peace
Original link: http://www.aljazeera.com/indepth/features/2017/05/scientist-race-build-peace-machine-170509112307430.html
Ambitious and inspiring project - to build a “Peace Machine” capable to ensure world peace. The idea feels to ignore many realities how wars are born, but helping people to communicate better, to understand each other’s semantics better sounds like a project worth doing.
Link to post
Posted on 2017 June 10 by polgarp · Tagged in Bias
Original link: https://hackernoon.com/algorithms-arent-racist-your-skin-is-just-too-dark-4ed31a7304b8
Since combating algorithmic bias seems one of the biggest challenge also for product design, “Algorithmic Justice League” is an awesome initiative.
Link to post
Posted on 2017 June 9 by polgarp · Tagged in Video, Kurzgesagt, and Overview
Original link: https://www.youtube.com/watch?v=WSKi8HfcxEk
While examining the proliferation of AI tech and it’s effect on society from a more broader perspective, as every Kurzgesagt video, this one is also pretty insightful.
Link to post
Posted on 2017 June 7 by polgarp · Tagged in Comics and Fun
Original link: http://www.commitstrip.com/en/2017/06/07/ai-inside/
“I knew it. It’s just ‘IFs’”
Link to post
Posted on 2017 June 4 by polgarp · Tagged in Google, Productivity, and Data visualization
Original link: https://blog.google/products/g-suite/visualize-data-instantly-machine-learning-google-sheets/
Nice addition to Sheets, it’s essentially just a natural language processor (still impressive). Since charts are very visual, selecting charts also seems to be a very visual task. So the question is how useful to hide chart selection behind a command line like interface.
Link to post
Posted on 2017 June 4 by polgarp · Tagged in Data collection
Original link: https://www.linkedin.com/pulse/training-ai-system-ray-garcia
Interesting take on the biggest issue in training a machine learning system: getting enough quality data.
Link to post
Posted on 2017 June 4 by polgarp · Tagged in Code generation and Design implementation
Original link: https://thenextweb.com/apps/2017/05/26/ai-raw-design-turn-source-code/
Interesting idea to generate front-end code based on mockups, however there are many similar design systems that can produce reasonable front-end code. What’s rather intriguing is the prospect of generating code that conforms to other standards, like is optimized for speed or can predict ideal micro animation.
Link to post
Posted on 2017 June 4 by polgarp · Tagged in Research and Interaction
Original link: https://arxiv.org/pdf/1705.07996.pdf
One of the challenges of designing AI products is the fundamental framework in which the interaction happens. The concepts presented in this paper reinforce the notion that is more like human-human interaction, meaning both parties need to be not just considered but tried to change, like make people more aware of their cognitive biases.
Link to post
Posted on 2017 June 4 by polgarp · Tagged in Statistics and Future AI
Original link: https://www.technologyreview.com/s/607970/experts-predict-when-artificial-intelligence-will-exceed-human-performance/
Some fun numbers from a research done two years ago. There are already some numbers that just didn’t work out as they were predicted (AI is now better in Go). Still amusing to see that AI researchers predict AI will be better in AI research very far in the future. I bet other experts would predict this differently.
Link to post
Posted on 2017 June 4 by polgarp · Tagged in IDEO, Design research, and Job automation
Original link: https://labs.ideo.com/2017/05/30/could-ai-replace-design-researchers/
Interesting design project, but even more interesting are the research experiment IDEO’s team did. There is certainly a big category of design research that’s currently being done via surveys and kind of fixed interviews that can be automated even via chatbots. I wonder if people are indeed more likely to tell things to an AI system, compared to a human, as building trust is usually a huge part of doing interviews.
Link to post
Posted on 2017 June 3 by polgarp · Tagged in AI development
Original link: https://freedom-to-tinker.com/2017/05/31/what-does-it-mean-to-ask-for-an-explainable-algorithm/
Working with AI tech one of the challenges is to understand what exactly is happening. This article is about a useful classification of different types of explanations, and this may hint on some ways it can be designed for.
Link to post
Posted on 2017 June 3 by polgarp · Tagged in Business model and Co-creation
Original link: http://www.nesta.org.uk/blog/err-algorithm-algorithmic-fallibility-and-economic-organisation
A fascinating post about the economic model of using algorithmic tools. Deciding what products should be designed with AI tech is a huge challenge, and this model gives a few pointers. Adding supervisors to check the accuracy seems to add another layer of complication, something else to design for (hello service design), but makes sense. Humans make different errors than computers, and end-users don’t expect computers to make different, so “dumb” mistakes.
Link to post
Posted on 2017 June 3 by polgarp · Tagged in Photography
Original link: https://www.theverge.com/circuitbreaker/2017/6/2/15728332/arsenal-dslr-camera-ai-helper-kickstarter
I was expecting photo-related products to work more after a picture was taken, but this makes also some sense. Although in a day when real images can be generated this may be more about the experience of taking pictures.
Link to post
Posted on 2017 June 3 by polgarp · Tagged in Nutella, Advertisement, and Algorithmic creativity
Original link: https://www.theverge.com/2017/6/2/15731648/nutella-packaging-algorithm-software
Maybe not an AI, still a fun use of algorithmic creativity.
Link to post
Posted on 2017 June 3 by polgarp · Tagged in Talk and Co-creation
Original link: https://www.ted.com/talks/garry_kasparov_don_t_fear_intelligent_machines_work_with_them
This is a nie talk by chess-champion Garry Kasparov who was beaten by IBM’s Deep Blue in 1997. He presented some ideas, on how computers and humans could work together, but there was one I liked the most. Better interaction is stronger than a great human and a great computer combined with a weak interaction. Something worth designing for.
Link to post
Posted on 2017 June 3 by polgarp · Tagged in Chatbot and Healthcare
Original link: https://medium.com/medilad/can-a-period-tracking-chatbot-pave-the-way-to-a-womens-health-ai-f6d35a0d4edc
There is a nice story here, how chatbots can evolve into more useful products in the future with AI tech.
Link to post
Posted on 2017 June 3 by polgarp · Tagged in Google and AlphaGo
Original link: https://www.theverge.com/2017/5/30/15712300/alphago-ai-humanity-google-artificial-intelligence-ke-jie
Nice report on AlphaGo’s current state and some interesting insights how Google’s engineers progress with it.
Link to post
Posted on 2017 May 30 by polgarp · Tagged in IBM, Amazon, and Emotional design
Original link: http://www.adweek.com/digital/why-tech-companies-like-ibm-and-amazon-brand-artificial-intelligence-with-human-names/
I’m surprised they don’t mention the most important reason - people have a harder time to relate to a thing than to a person, so naming also makes these systems (and any AI system for that matter) more person like. This also makes the mistakes they make more forgivable. With AI products it’s imperative to have the right emotional design.
Link to post
Posted on 2017 May 30 by polgarp · Tagged in AI development and Healthcare
Original link: https://qz.com/989137/when-a-robot-ai-doctor-misdiagnoses-you-whos-to-blame/
In non-AI medical products the manufacturer can clearly take the blame, design flaws can be discovered after an incident. Learning technology learns from available data too, so data mistakes can make the designers responsible. And the there is also the point of: learning systems make different mistakes than humans. If the doctor does the diagnosis, even with the support of an AI assistant it’s still his call. So I’m guessing it’s up to the design of the system, at minimum it has to inform the doctor on the strength of it’s assessment.
Link to post
Posted on 2017 May 30 by polgarp · Tagged in Legal, Overview, and Job automation
Original link: http://www.topbots.com/automating-the-law-a-landscape-of-legal-a-i-solutions/
Interesting overview on AI tech products from the legal industry. It seems most of these products automate a few specific workflows that were maybe easy to spot, as something screaming for automation. I wonder what are the biggest design challenges here - besides building trust in a tech-sceptic industry that is.
Link to post
Posted on 2017 May 29 by polgarp · Tagged in Healthcare and Overview
Original link: https://medium.com/@medhaa/radical-approaches-to-healthcare-our-thesis-in-computational-biology-ee19f4f3da31
Nice overview mostly on the healthcare market opportunities for AI tech, and I’m guessing just overall better designer services.
Link to post
Posted on 2017 May 29 by polgarp · Tagged in AI chips
Original link: https://www.theverge.com/2017/5/29/15707606/arm-cortex-a75-a55-mali-g72-specs-announced
So I was mistaken that only Apple’s devices will have initially powerful on device AI teach capabilities. ARM seems to be in the game too.
Link to post
Posted on 2017 May 28 by polgarp · Tagged in Google, Music making, and Co-creation
Original link: https://www.fastcompany.com/40424051/how-googles-music-making-ai-learns-from-human-minds-at-festivals
This feels like a pretty exciting application, but also fits into the broader theme of making people more creative with AI tools. What’s also interesting here is the way these engineers approach the development, they are essentially user testing their app. I wonder why they didn’t approach Google’s design team.
Link to post
Posted on 2017 May 28 by polgarp · Tagged in IBM and Microsoft
Original link: https://www.youtube.com/watch?v=rw4aWeL3v5g
IBM + Microsoft CEO, MIT head - quite general discussion how AI tech might transform businesses and the broader society. Some interesting points, but nothing new if you follow tech news anyway.
Link to post
Posted on 2017 May 27 by polgarp · Tagged in Apple and AI chips
Original link: https://www.theverge.com/2017/5/26/15702248/apple-neural-engine-ai-chip-iphone-ipad
Maybe just another Apple product gossip, maybe not. But Apple does have the capability to create their own chip. Probably Google has something similar in the works too, but what other Android OEMs have the same capability, or ready to license the tech from somebody? As we assume AI assistants will make the users smarter, this would mean iPhone users will be smarter (and high end Android users too).
Link to post
Posted on 2017 May 27 by polgarp · Tagged in Google, Go, and Co-creation
Original link: https://blog.google/topics/google-asia/alphagos-next-move/
I would assume that a game like go has such a high level of sophistication that it’s a huge effort to develop new strategies. The big news is not that a computer can now beat the best go players, but how it can inspire and enhance player capabilities and strategies. This should be the minimum goal for any product designed with AI tech.
Link to post
Posted on 2017 May 25 by polgarp · Tagged in Google and Journalism
Original link: https://blog.google/topics/journalism-news/ai-newsroom-whats-happening-and-whats-next/
Of course AI tech is bound to change more than one industry, but news seems like a special beast.
Link to post
Posted on 2017 May 23 by polgarp · Tagged in Management and Training
Original link: http://www.work-bench.com/blog/2017/05/17/work-bench-perspectives-ai-for-next-generation-managers/
Just a little idea nugget on how AI tech could impact manager training. I talk frequently about how management can be designed the same way as any other product - even if some method details differ. So if it’s like any other products, the same tech can be used.
Link to post
Posted on 2017 May 23 by polgarp · Tagged in Job automation
Original link: https://www.washingtonpost.com/news/wonk/wp/2017/05/15/the-worst-paying-jobs-for-college-grads-boast-this-sneaky-advantage/
I don’t like these type of articles, as they focus too much on the “replacement” part, not how AI tech can make every workers life easier by making them smarter - this goes even for teachers and social workers. Especially replacing software engineers are too far fetched, there are too many realities of development this article ignores. But the fears are there, so design’s role is crucial, as holistic product view has to consider the broader impact of deployed systems.
Link to post
Posted on 2017 May 23 by polgarp · Tagged in Ethics
Original link: https://medium.com/@sebastianjones_95354/4aab445873c8
Still a draft article, but still interesting on ethics role in AI products.
Link to post
Posted on 2017 May 23 by polgarp · Tagged in Ambient AI
Original link: https://motherboard.vice.com/en_us/article/how-ai-could-make-urban-bicycling-a-whole-lot-easier
This idea fits well into the ambient AI perspective, not just specific technological products, but everything around us can get better.
Link to post
Posted on 2017 May 23 by polgarp · Tagged in Google and Strategy
Original link: https://stratechery.com/2017/boring-google/
An other take on the announcements of Google I/O, more from the product perspective.
Link to post
Posted on 2017 May 21 by polgarp · Tagged in Talk and Principles
Original link: https://www.ted.com/talks/stuart_russell_how_ai_might_make_us_better_people
Almost like Asimov’s laws, in approach certainly.
Link to post
Posted on 2017 May 21 by polgarp · Tagged in Comics and Fun
Original link: https://xkcd.com/1838/
XKCD on ML.
Link to post
Posted on 2017 May 21 by polgarp · Tagged in Job automation and Training
Original link: https://qz.com/979812/how-ai-will-change-the-shape-of-organizations/
I don’t think “doing the grunt work” is the best way to learn, it was just the most cost efficient way to do so. If grunt work is done by AI tech, we can design better learning experiences. This also shows why thinking in systems is important, when automatizing certain workflows, we can design learning systems in parallel to keep all the knowledge inside the organization.
Link to post
Posted on 2017 May 21 by polgarp · Tagged in Security
Original link: https://hbr.org/2017/05/ai-is-the-future-of-cybersecurity-for-better-and-for-worse
As designers we are usually not that concerned with adversarial use of the systems we are building, since most possible harm tends to be on financial level, and we just kinda expect the engineers to handle everything. This will change if the attackers are able to change the perception of reality, much more than current phishing attacks. Seems challenging to design to.
Link to post
Posted on 2017 May 20 by polgarp · Tagged in Google and Apple
Original link: https://www.fastcompany.com/40422573/why-apple-is-watching-googles-ai-progress-carefully
“Apple doesn’t need to demonstrate superiority over Google in the pure science part of artificial intelligence. It can win by coupling its AI chops with a superior understanding of how people will best benefit from AI in day-to-day life.”
Link to post
Posted on 2017 May 20 by polgarp · Tagged in Google and Assistant
Original link: https://www.theverge.com/2017/5/17/15651246/google-assistant-iphone-ai-future-interface-io-2017
Great analysis on Google’s Assistant product (plenty of new features of it were just announced at Google I/O). While UX seems indeed to be a driving force here, I’m not so sure if it can be compared to previous UI paradigm shifts. Also while UX is indeed a huge challenge, product questions are as many (Google has to work across a huge portfolio of products, and probably also integrate with third parties), to best UX won’t work without a great product and vice-versa.
Link to post
Posted on 2017 May 20 by polgarp · Tagged in Neuralink
Original link: https://medium.com/startup-grind/the-implications-of-neuralink-1409fa277487
Some very interesting predictions, however I’d disagree with some of it. The author makes the same mistake as many others, thinking that Apple’s design advantage is the user interface. This is misunderstanding on what design is capable of - and more importantly how much design we will need in the future (I think a lot).
Link to post
Posted on 2017 May 20 by polgarp · Tagged in Google
Original link: https://blog.google/topics/machine-learning/making-ai-work-for-everyone/
Plenty of interesting announcements on this year’s Google I/O by Google CEO Sunder Pichai. While the design of individual products is pretty exciting in itself, the technologies Google’s making available is even more awesome.
Link to post
Posted on 2017 May 20 by polgarp · Tagged in Google and AI chips
Original link: https://www.theverge.com/2017/5/17/15645908/google-ai-tensorflowlite-machine-learning-announcement-io-2017
Great for a few reasons - little loading time (if connectivity is slow), opens up new possibilities in privacy related apps, results is even more personal. But at the same time will consume more battery, and the overall learning needs to be even better optimized to be fast enough. Which means plenty of design will be needed around the algorithm.
Link to post
Posted on 2017 May 20 by polgarp · Tagged in Google
Original link: https://www.theverge.com/2017/5/19/15660610/google-photos-ai-relationship-emotional-labor
After improving our mental labor I think it’s the logical next step for technology to also improve our emotional labor by automatizing some of the more mundane tasks.
Link to post
Posted on 2017 May 20 by polgarp · Tagged in Google, Google Lens, and Computer vision
Original link: https://www.recode.net/2017/5/19/15666704/google-lens-key-example-ai-first-computer-vision
Still feels like just a bunch of cool things that Google can do with AI, and not a frequent, everyday problem, but at least Google Lens does show the awkwardness of voice UI’s, barely better then typing into a search box.
Link to post
Posted on 2017 May 20 by polgarp · Tagged in Facebook and Chatbot
Original link: https://www.theverge.com/platform/amp/2017/5/15/15640886/facebook-parlai-chatbot-research-ai-chatbot
“… one of the objectives of this is to have your own digital friend.” Just as we live in the future and imagination is not a limit anymore, our imagined friends from childhood can also come to life.
Link to post
Posted on 2017 May 20 by polgarp · Tagged in Co-creation
Original link: https://www.theverge.com/2017/5/17/15654350/bots-error-noise-collective-behavior-decision-making-science-puzzle
Intriguing results. One huge challenge seems in designing systems with AI is the interaction between the human and machine actors. While the assumption is that people can be smarter when working with AI, it’s maybe not that easy than to just create more clever algorithms.
Link to post
Posted on 2017 May 18 by polgarp · Tagged in Overview
Original link: https://medium.com/@pedromdd/ten-myths-about-machine-learning-d888b48334a3
This one is for the wider audience clarifying some of the common sentences the media is writing about machine learning.
Link to post
Posted on 2017 May 18 by polgarp · Tagged in Google
Original link: https://google.ai/
Super interesting for product people and probably devs, but also undeniable Google to put technology this front.
Link to post
Posted on 2017 May 18 by polgarp · Tagged in Apple
Original link: https://techcrunch.com/2017/05/13/apple-acquires-ai-company-lattice-data-a-specialist-in-unstructured-dark-data/
Just another sign that Apple is actually doing something in the AI space.
Link to post
Posted on 2017 May 14 by polgarp · Tagged in Google, Allo, and Illustration
Original link: https://research.googleblog.com/2017/05/neural-network-generated-illustrations.html
Not sure how much it adds to have your selfies converted into stylized avatars, but it does look like fun. The design of the customization possibilities is a remarkable feature however. It enables the users to personalize the results - even if the end result was perfectly spot on, some customization helps to own the avatar.
Link to post
Posted on 2017 May 14 by polgarp · Tagged in IBM and Voice recognition
Original link: https://www.theverge.com/2017/5/11/15621930/star-trek-bridge-crew-voice-commands-ibm-watson
Nice. Soon Scottie doesn’t need to speak to the mouse anymore.
Link to post
Posted on 2017 May 14 by polgarp · Tagged in Book
Original link: http://rosenfeldmedia.com/books/designing-agentive-technology/
New Rosenfeld book on agentive technologies by Christopher Noessel. “Advances in narrow artificial intelligence make possible agentive systems that do things directly for their users (like, say, an automatic pet feeder). They deliver on the promise of user-centered design, but present fresh challenges in understanding their unique promises and pitfalls. Designing Agentive Technology provides both a conceptual grounding and practical advice to unlock agentive technology’s massive potential.”
Link to post
Posted on 2017 May 14 by polgarp · Tagged in Design
Original link: https://medium.com/startup-grind/design-in-an-age-of-artificial-intelligence-739e656b44ba
Design jobs, just as any white collar job will change as a result of rise of AI tech. But I think instead of “how we will work”, a much better question is how we can create better products for users with these tech. One thing I agree though - “designers should code” will get old fast, prepare for “designers have to write algorithms”.
Link to post
Posted on 2017 May 14 by polgarp · Tagged in Job automation and Statistics
Original link: https://www.recode.net/2017/5/10/15612570/automation-ai-acceleration-revenue-growth-financial-upside-mnuchin
So automation makes companies more productive, contributes to growth, so they can hire more. While companies not automating will slowly shrink, not because of automating, but because they haven’t improve their productivity. Same as becoming more design focused.
Link to post
Posted on 2017 May 14 by polgarp · Tagged in Overview and Source
Original link: http://aiplaybook.a16z.com/
Awesome introduction / overview / source on all topics AI by the VC firm Andreessen Horowitz. While not created for designers, it contains tons of useful stuff.
Link to post
Posted on 2017 May 9 by polgarp · Tagged in Overview and Product
Original link: https://hackernoon.com/a-list-of-artificial-intelligence-tools-you-can-use-today-for-personal-use-1-3-7f1b60b6c94f
A nice list of products with AI tech, later part of the series also has a list of B2B and other industry products.
Link to post
Posted on 2017 May 8 by polgarp · Tagged in Ethics
Original link: https://www.technologyreview.com/s/604270/real-or-fake-ai-is-making-it-very-hard-to-know/
With such tech, reality is the new virtual reality. Similar to UX dark patterns, maybe using something like this will be considered dark use. But also the question will be, how can we teach users and ultimately educate people as the lines between facts and fake blur.
Link to post
Posted on 2017 May 6 by polgarp · Tagged in Business models and Product management
Original link: https://news.greylock.com/the-new-moats-53f61aeac2d9
Fascinating read on business models, and how companies can create “moats”, defensible business positions from competitors. While AI devaluates existing advantages it is capable to add new ones. As design is adding more and more value to products, thinking about overall direction and vision of the product is must for all designers. Whenever you create a new product or new features, ask your product manager how it enhances your “business moats”.
Link to post
Posted on 2017 May 6 by polgarp · Tagged in Fun and Lego
Original link: https://jacquesmattheij.com/sorting-two-metric-tons-of-lego
Half fun, half awesome business idea.
Link to post
Posted on 2017 May 6 by polgarp · Tagged in Job automatization
Original link: https://www.nytimes.com/2017/04/28/technology/meet-the-people-who-train-the-robots-to-do-their-own-jobs.html
Remarkable read with quite interesting cases. One of them is the interaction designer for the assistant company x.ai. Her example shows how much work is needed to make an assistant with AI tech actually behave acceptable for users, a work for that designers are perfectly matched
Link to post
Posted on 2017 May 6 by polgarp · Tagged in Google, Product, and Productivity
Original link: https://blog.google/products/g-suite/how-machine-learning-g-suite-makes-people-more-productive/
I really like the basic premise here: make mundane tasks easier with machine learning, to allow more creative time for office workers. This is something UX design is also striving for, to make workflows more efficient and more enjoyable. But besides the smart reply feature from Inbox they rather look like just an added feature to the interface and not something designed with the core workflow in mind.
Link to post
Posted on 2017 May 6 by polgarp · Tagged in Emotions
Original link: https://iq.intel.com/how-ai-and-deep-learning-help-explain-human-fear/
Designing for emotions is a powerful approach. Even negative emotions, like fear, give designers both an opportunity and a challenge. Having an emotional understanding in AI tech is also both - an opportunity to create more humanistic products and also a challenge to create them right
Link to post
Posted on 2017 May 6 by polgarp · Tagged in Robotics and Computer vision
Original link: https://www.theverge.com/2017/5/4/15544118/prosthetic-bionic-hand-arm-ai-object-recognition
Another nice example how AI tech can make it way cheaper and simpler to solve basic human needs in a better way, in this case helping prosthetic wearers grab things. Makes many categories of design problems easier to solve.
Link to post
Posted on 2017 May 6 by polgarp · Tagged in Enterprise and Product management
Original link: https://twitter.com/saranormous/status/857440599221219328
Sarah Guo’s tweetstorm is super interesting, especially if you are designing business products with AI tech. Plenty of opportunity for great design - but also caution, having great AI tech in itself is far from enough, unless your product is also capable to change workflows, and your team able to teach customers. This can only be done through extensive user research and careful design.
Link to post
Posted on 2017 April 30 by polgarp · Tagged in Superintelligence
Original link: https://backchannel.com/the-myth-of-a-superhuman-ai-59282b686c62
When talking about the future of AI, it’s inevitable that the conversation shifts to machines becoming smarter than humans, maybe even in the near future. This article describes some of the assumptions behind this belief in light of our current knowledge.
Link to post
Posted on 2017 April 30 by polgarp · Tagged in Personalization
Original link: https://apnews.com/ab2742712f314257991c5598ea98fa16/Icelandic-language-at-risk;-robots,-computers-can't-grasp-it
While not strictly AI news, the disappearing Icelandic language shows challenges with the promise of current AI tech, generally trained algorithms rarely include even fairly large chunks of human culture, like a language spoken by 400,000 people. This compares to the deep personalization promise of many products.
Link to post
Posted on 2017 April 30 by polgarp · Tagged in Dropbox, Computer vision, and OCR
Original link: https://blogs.dropbox.com/tech/2017/04/creating-a-modern-ocr-pipeline-using-computer-vision-and-deep-learning/
Fascinating read on Dropbox’s new OCR system with lots of juicy details on how they ended up with their final product. While the product goal was clear, there were lots of human issues besides the engineering challenges - like patterns in data of the real world use needed to be understood, and even the training data collection tool needed to be designed. Some good tips if you are working on something similar.
Link to post
Posted on 2017 April 30 by polgarp · Tagged in Design patterns and Overview
Original link: https://www.smashingmagazine.com/2017/04/applications-machine-learning-designers/
This is a good overview on how designers can think about current AI capabilities to include in their products, and there are also a few design patterns.
Link to post
Posted on 2017 April 29 by polgarp · Tagged in AI chip
Original link: https://www.wired.com/2017/04/race-make-ai-chips-everything-heating-fast
First you will have AI chips on the server / cloud side, with GPUs on the device level, later maybe integrated, low-power chips in the device? Seems to be a requirement for designing systems with local training - that is fast enough for consumer use.
Link to post
Posted on 2017 April 29 by polgarp · Tagged in Facebook
Original link: https://www.buzzfeed.com/alexkantrowitz/meet-the-man-who-makes-facebooks-machines-think
Nice report on Facebook’s leading AI researcher, Yann LeCun. It’s interesting to see how far ahead their research goes, and also their decisions on some of the products they are capable to do (e.g. fake news filtering), but decide to not do.
Link to post
Posted on 2017 April 29 by polgarp · Tagged in Product and Assistant
Original link: https://gradeproof.com/
Getting phrasing suggestions in addition to spelling and grammar seems pretty useful. This could work in small, unobtrusive ways, even automatically fixing errors. Still, it’s just fixing a minor annoyance in life, and not necessarily make the user actually write better.
Link to post
Posted on 2017 April 29 by polgarp · Tagged in Google and Speech recognition
Original link: https://www.recode.net/2017/4/18/15345740/google-speech-recognition-ai-cloud
AI tech is getting commoditized fast, devs can almost just plug in. So the discussions should soon focus from “can we implement it” to “how we should implement it”. Great times ahead for both data scientists and designers.
Link to post
Posted on 2017 April 29 by polgarp · Tagged in Bias and Ethics
Original link: https://techcrunch.com/2017/04/25/faceapp-apologises-for-building-a-racist-ai/
Feels like AI tech is not making product biases worse, it just highlights more if an organization has such a problem. Biased organizations will design biased products.
Link to post
Posted on 2017 April 29 by polgarp · Tagged in Ethics and Justice system
Original link: https://www.wired.com/2017/04/courts-using-ai-sentence-criminals-must-stop-now/
Using algorithms in criminal justice shows on hand how this tech can enhance white collar work instead of just replacing them. On the other hand I like the idea of a similar approval system for algorithmic tools impacting human lives than there are in place for drugs and medical devices. Maybe it could show the weakness of systems trained with bad / incomplete data.
Link to post
Posted on 2017 April 29 by polgarp · Tagged in Amazon and Product
Original link: http://www.theverge.com/2017/4/27/15447834/amazons-echo-look-ai-analysis-concerns
Having somebody tell you if you look sharp or not is an interesting application, even if this somebody is a shop clerk who just wants to sell you more clothes in the end and is not that concerned with you personally. Since this tech is capable to find out so much more about you - that maybe even you don’t know about yourself, trust as a product / brand feature will be even more important in the future. You wouldn’t trust just anybody to look around in your closet.
Link to post
Posted on 2017 April 28 by polgarp · Tagged in Fonts and Experiment
Original link: http://fontmap.ideo.com/
More fun than useful, IDEA mapped 750 fonts via machine learning to explore connections. Machine learning is bound to change knowledge / intuition dependent tools and workflows, even for designers.
Link to post
Posted on 2017 April 28 by polgarp · Tagged in Service
Original link: https://deepai.host/
A new machine learning as a service app.
Link to post
Posted on 2017 April 28 by polgarp · Tagged in Product and Email
Original link: https://www.helloastro.com/
If anything, email definitely needs a huge AI upgrade. Combining your email app with a learning bot seems like an interesting idea. If at the least it manages to remind of important messages it’s a step forward.
Link to post
Posted on 2017 April 22 by polgarp · Tagged in Neuralink and Brain interface
Original link: http://waitbutwhy.com/2017/04/neuralink.html
Long, but extremely fascinating read on Elon Musk’s new venture by Tim Urban. While the whole topic should be of general interest to anybody looking in the future, the idea of having AI systems as a kind of brain extension sounds to me a good general principle when creating such systems.
Link to post
Posted on 2017 April 22 by polgarp · Tagged in Ethics and Art
Original link: https://worldwritable.com/ethical-imperatives-in-ai-and-generative-art-b8cf51af4c5
AI design principles have to do a lot with the ethical standard of the people actually designing it. And since we are designing for people, there is a lot of misuse (intended or unintended) possible. This article provides a few interesting point, for example data sets provided by people can be intentionally altered making it in the best case a minor annoyance, in the worst case a critical error.
Link to post
Posted on 2017 April 22 by polgarp · Tagged in Talk and Overview
Original link: https://www.youtube.com/watch?v=g74Gprwyvf8
This talk overviews some more general AI topics, and goes just a little bit deeper. No direct design connection, but still interesting.
Link to post
Posted on 2017 April 21 by polgarp · Tagged in Product and Assistant
Original link: https://chrome.google.com/webstore/detail/writefull-thesaurus/gjaljmejoplhicfhppcfcgjojehddfcc?hl=en
This is how I imagine a great product built with AI tech: it just makes me overall more awesome. This one helps me to write better - and not being arrogantly pushy while doing it.
Link to post
Posted on 2017 April 20 by polgarp · Tagged in Development
Original link: https://www.technologyreview.com/s/604087/the-dark-secret-at-the-heart-of-ai/
“We don’t really know how it does what it does” is troubling. Even if we will have systems capable of explaining how they came to a conclusion, for AI tech to be seamlessly integrated into everyday applications, they should mostly follow people’s mental models and shouldn’t act surprising. This is a basic interaction principle. A self driving car hitting a tree randomly is definitely surprising.
Link to post
Posted on 2017 April 19 by polgarp · Tagged in Product, iOS, and App
Original link: https://itunes.apple.com/app/id1225427550
An AI assisted flowchart drawing tool. Seems these kind of small apps - basic functionality enhanced with AI that results in a better experience overall just start to pop out as AI tech gets easier to implement and design.
Link to post
Posted on 2017 April 17 by polgarp · Tagged in Mattels, Product, and Assistant
Original link: https://www.fastcompany.com/40400777/mattels-new-ai-will-help-raise-your-kids
Having a voice assistant that answers all of your kid’s “Why” questions actually sounds pretty useful. I wonder if a design is possible where you also spend more quality time with your kid.
Link to post
Posted on 2017 April 17 by polgarp · Tagged in Google, Experiments, and Collection
Original link: https://aiexperiments.withgoogle.com/
A listing site for all the interesting stuff Google has been doing in the AI space. Probably worth to keep an eye on as new experiments pop up, but you can also submit your own. Such tinkering definitely can lead to a better understanding on how humans can interact with AI tech.
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Product and Analytics
Original link: https://voiceops.com/
This app takes the phone calls of sales agents, transcribes them, and provides detailed analytics to their managers. Interesting to see that the machine learning in the voice transcription and the analytics section provides data to the organization’s learning.
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Google, Research, and Drawing
Original link: https://research.googleblog.com/2017/04/teaching-machines-to-draw.html
Google researcher’s latest drawing model, sketch-rnn. There are lots of applications of an automatic drawing system besides the obvious ones suggested by the authors (pattern generation and “sketch autosuggest”). Also kind of takes out the art of the more illustrative designs.
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Course and Natural language processing
Original link: https://www.youtube.com/playlist?list=PL3FW7Lu3i5Jsnh1rnUwq_TcylNr7EkRe6
Stanford University published it’s lecture series about NLP. Sounds like an excellent intro into the topic.
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Image recognition and Attack
Original link: http://www.theverge.com/2017/4/12/15271874/ai-adversarial-images-fooling-attacks-artificial-intelligence
Fooling machine learning systems with specifically crafted content so adversarial attacks should be the concern of any system designer not just security people. There are also cases you don’t want systems to recognize your face, to maintain privacy from automatic systems (like Facebook tagging you on pictures).
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Startups
Original link: https://medium.com/money-talks-the-official-abe-blog/how-to-fail-with-artificial-intelligence-b3c4b1966bb3
Sounds a bit funny description on how startups working with AI can fail, but these points are true for startups working with any technology. I guess the deeper point here is that product thinking and design is needed even if you have the most awesomely trained model.
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Google, Research, and Mobile
Original link: https://research.googleblog.com/2017/04/federated-learning-collaborative.html
Teaching models on the mobile’s, why not forwarding private data into the cloud sounds like something Apple would want to do, so it’s interesting Google is publishing a way to do this. This also sounds useful for general product development - you can release your app even without a properly trained network and / or enough user data, as the training can happen while the user’s are already using your app.
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Google, Research, Drawing, and Toy
Original link: https://www.blog.google/topics/machine-learning/fast-drawing-everyone/
A nice toy by Google based on their sketching research, that autosuggest images based on the user’s scribbles. Although I didn’t manage to generate a human form from my stick figures, it’s quite impressive. Sounds like a great addition into an image search tool.
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Ethics and Personalization
Original link: https://www.ted.com/talks/eli_pariser_beware_online_filter_bubbles
An older TED talk by Eli Pariser about the filter bubbles personalization and algorithmic editing is creating. I loved his comparison on dietary needs, just like with food, you have to take care what you consume, you cannot go always for chocolate cake, you need to eat also vegetables - looking out of your filter bubble. Good idea for designers to think about food labels applied to online content, fighting fake news anyone?
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Job automatization
Original link: https://machinelearnings.co/how-to-prepare-your-career-for-artificial-intelligence-driven-automation-1bb153759b3b
Excellent article on how automation will affect jobs, and what anyone can do about it (spoiler: learn more and try to be a better human). I think it also highlights well, that designing systems with AI tech needs to be really well done. Not just to harness the projected productivity increase, but also for the automated systems play well with their human contacts.
Link to post
Posted on 2017 April 15 by polgarp · Tagged in Design and Overview
Original link: https://www.oreilly.com/ideas/ai-and-the-future-of-design-what-will-the-designer-of-2025-look-like
I find the theme of this article a bit pessimistic, although I agree with it’s main conclusion: interaction design of AI systems will be an interesting and challenging field, as they spread.
Link to post
Posted on 2017 April 14 by polgarp · Tagged in Assistant, Chatbot, and Legal
Original link: https://www.theverge.com/2017/5/10/15609850/yoshua-bengio-bolter-ai-chatbot-legal-advice-immigration
Offering legal advice and assistance has been the domain of lawyers, a complicated field that is easy to automate with the correct tech. In essence a translation service from normal human language to the nuanced and exact formulation of laws. This however assumes the system of policies form a coherent and logical whole - which is not necessary true of any system designed and built by humans.
Link to post
Posted on 2017 April 8 by polgarp · Tagged in Fun, Neural network, and Dating
Original link: http://lewisandquark.tumblr.com/post/159302925452/the-neural-network-generated-pickup-lines-that-are
“Hello.” sounds like a winning one.
Link to post
Posted on 2017 April 8 by polgarp · Tagged in Fun, Neural network, and Cooking
Original link: http://lewisandquark.tumblr.com/post/140219420017/the-silicon-gourmet-training-a-neural-network-to
Sounds like fun to create a restaurant based around automatically generated recipes.
Link to post
Posted on 2017 April 8 by polgarp · Tagged in Management and Enterprise
Original link: https://hbr.org/2017/03/how-pg-and-american-express-are-approaching-ai
Organizations create products that mirror themselves, so it’s interesting how enterprises set up their groups working with AI. Customer benefit is where the design magic happens, and much of the clerical work behind support can be improved with AI tech. How the customers will feel better treated will be challenge.
Link to post
Posted on 2017 April 8 by polgarp · Tagged in Facebook, Assistant, and Chatbot
Original link: https://www.recode.net/2017/4/6/15203526/facebook-messenger-m-artificial-intelligence-ai-bots
Having an assistant helping you to navigate the daily life, like Alfred to Bruce Wayne seems like an awesome idea. But even if you use one messaging tool it doesn’t contain your whole life, and rudimentary help may come over as too pushy. It feels hard to design something that is emphatic enough.
Link to post
Posted on 2017 April 8 by polgarp · Tagged in Development and Debug
Original link: http://asia.nikkei.com/Tech-Science/Tech/Ben-Horowitz-worries-about-who-will-debug-artificial-intelligence
Interesting take on a more practical issue in AI tech. Most bugs affect the UX and many times the push for fixing experience degrading bugs falls to the people doing the usability tests. So what will happen if we don’t even understand the nature of the bug.
Link to post
Posted on 2017 April 7 by polgarp · Tagged in Resources, Design, and Algorithmic design
Original link: http://algorithms.design/
Awesome collection of articles and tools on how artificial intelligence impacts design by Yury Vetrov.
Link to post
Posted on 2017 April 4 by polgarp · Tagged in Voice UI and Analysis
Original link: http://ben-evans.com/benedictevans/2017/2/22/voice-and-the-uncanny-valley-of-ai
Great article on the issues of voice UIs by Benedict Evans. While the interaction style seems new, it’s very similar to classic command line: the users have to remember what they can do, even if they don’t need to formulate their commands exactly. The smarter the assistant appears the more likely frustration happen due to not smart enough.
Link to post
Posted on 2017 April 4 by polgarp · Tagged in Research, Adversarial neural networks, and Computer vision
Original link: https://arxiv.org/pdf/1610.08401v1.pdf
Just as the spreading of web ads gave birth to ad-blockers, most spreading of tech results in counter tech. Seems like if you are disturbed by ever present systems recognizing your images, it’s pretty easy to confuse them.
Link to post
Posted on 2017 April 4 by polgarp · Tagged in Talk and Ethics
Original link: https://www.youtube.com/watch?v=WjKdKvDS10g
Nice talk on ethics for algorithms by Abe Gong from last year. Ethical algorithms stand pretty close to systems designed according to user’s needs.
Link to post
Posted on 2017 April 4 by polgarp · Tagged in StarCraft, Research, and Gaming
Original link: http://www.theverge.com/2017/4/3/15164490/alibaba-ai-starcraft-combat
Interesting how these agents were capable to learn and deploy tactics only the best players are able - and probably they will be soon even better, at the least they don’t have to deal with the physical limits of unit control.
Link to post
Posted on 2017 April 3 by polgarp · Tagged in Research, Medical, Diagnosis, and Labelling
Original link: https://arxiv.org/abs/1703.08774
Raising data quality results in better outcomes - just like with people: learning from multiple sources leads to a more nuanced knowledge. I wonder how the same weighting idea could improve recommendation systems.
Link to post
Posted on 2017 April 3 by polgarp · Tagged in Robots, Manufacturing, and Job automatization
Original link: https://www.recode.net/2017/4/3/15123006/robots-sold-america-growth-300-percent-jobs-automation
Robots taking over blue-collar jobs is nothing new, but they seem to reach mainstream now - probably that is why news about it seems to be shocked. This will happen to white-collar workers to in a few years.
Link to post
Posted on 2017 April 2 by polgarp · Tagged in Voice UI and Overview
Original link: https://www.cooper.com/journal/2017/3/this-is-the-year-of-voice-ui
Posted on 2017 April 2 by polgarp · Tagged in Art, Creativity, and Overview
Original link: https://www.viewpointmag.com/2017/03/28/the-darkness-at-the-end-of-the-tunnel-artificial-intelligence-and-neoreaction/
Posted on 2017 April 2 by polgarp · Tagged in Robots and Job automatization
Original link: https://www.fastcompany.com/3069269/robots-are-going-to-kill-jobs-because-they-already-have
Posted on 2017 April 2 by polgarp · Tagged in Job automatization
Original link: https://www.fastcompany.com/3069242/how-this-company-combines-the-gig-economy-model-with-ai-to-be-more-produc
Posted on 2017 April 2 by polgarp · Tagged in Google and Investment
Original link: https://9to5google.com/2017/03/31/google-brain-toronto-ai-deep-learning/
Posted on 2017 April 2 by polgarp · Tagged in Elon Musk, AI apocalypse, and Overview
Original link: https://www.vanityfair.com/news/2017/03/elon-musk-billion-dollar-crusade-to-stop-ai-space-x
Posted on 2017 April 2 by polgarp · Tagged in Elon Musk and Neuralink
Original link: https://www.theverge.com/2017/3/27/15077864/elon-musk-neuralink-brain-computer-interface-ai-cyborgs
Posted on 2017 April 2 by polgarp · Tagged in Pizza and Robots
Original link: https://www.recode.net/2017/3/29/15100748/dominos-deliver-pizza-robots-germany-starship
Posted on 2017 April 2 by polgarp · Tagged in Adobe, Photography, and Deep learning
Original link: https://www.theverge.com/2017/3/30/15124466/ai-photo-style-transfer-deep-neural-nets-adobe
Posted on 2017 April 2 by polgarp · Tagged in Amazon, Emotional design, and Assistant
Original link: http://digg.com/2017/amazon-alexa-is-not-your-friend
Posted on 2017 April 2 by polgarp · Tagged in Healthcare
Original link: http://www.newyorker.com/magazine/2017/04/03/ai-versus-md
Posted on 2017 April 2 by polgarp · Tagged in Social media and Bias
Original link: https://motherboard.vice.com/en_us/article/wnkbbw/ai-can-now-identify-racist-code-words-on-social-media