Hello, Dear Readers! My name is Karolina, and since May I have been working with Saidot on the Citizen Trust Through AI Transparency Project. It is also a base fore my Master Thesis in Human-Computer Interaction, titled: “Designing guidelines for trustworthy AI systems in the public sector”. My main job at Saidot is to lead the interactions with citizens and creating content for guidelines.
Ethics, here transparency, wasn’t something of my interest at the beginning of my studies. In fact, during my bachelor studies in Control Engineering and Robotics ethics was mentioned once or twice, but never as something important. It took me a few years and talks to make me understand that lack of discussions about ethics among students, who will soon be responsible for creating new products, is a problem. Therefore, I was more than happy to meet Meeri, Saidot's CEO, after her speech about Ethical AI and join the company forces for the Citizen project.
On the second week of May, I attended the annual Google conference, called Google I/O, invited as the Women Techmakers scholar. Let me keep it clear: I won’t cover all the points of the conference in the below text - for that you can check eg. articles from The Verge or TechRadar, or watch a vlog of other Women Techmaker scholar. Instead, I’ll touch upon topics that were particularly interesting for me - and hopefully will be of interest for you, dear reader. Enjoy!
Accessibility & inclusion in tech
It is good to see that Google works towards accessibility and inclusion in their products. This year, the most commonly repeated story was about Dimitri Kanevski, a research scientist at Google. In short, as he learnt English only after becoming deaf, his strong Russian accent made his speech become hardly recognisable by Google Assistant (GA). When realising this, he sat with another researcher, recorded a huge amount of phrases, plugged it into GA data and voilà - GA started recognising what Dimitri was asking. Now, Google asks anyone who has an unclear way of speech to record their samples in order to improve the general voice recognition; you can find more information at the official project website. The project doesn’t end at that, watch this video on how else they are enabling people to communicate:).
Here a brief list of some of the lessons you could take out of the talks and workshops presented at the conference:
1. Have in mind that majority of people around the globe cannot afford the newest smartphones. Yet, it is important that they can use basic tools like translators, web browsing etc. It is worth to work on compressing applications, so that they can be used on phones with little RAM and weak connection to internet.
2. Always, try to visualise the future users of your product and try to step in their shoes. Include people of different disabilities in the group. Note, that some some disabilities might be reflected in our day to day situations: e.g. travelling in the shaky train, getting sun-blind, carrying a child in one arm etc. How could you use the product in such situations?
For more information, watch Designing for Accessibility talk.
Ethical and fair AI & ML
is is heavily connected to it. In fact, if you asked me for that one talk I would recommend you to watch, I would point you towards Machine Learning Fairness: Lessons Learned speech. The presentation was very clear, interesting and full of good examples and lessons, such as:
a) accept that it will take longer to finish the project when working with ethical requirements;
b) make sure that everyone in the team is satisfied with doing a product that will comply with ethical norms;
c) remember, that ethics is very much dependent on geographical context.
Two examples provided on ML Fairness talk. Do you understand the need of working on that issue?
Collaboration between artists and AI
Something that I hadn’t expected, was attending so many art-related talks at Google I/O. Don’t get me wrong, I do like art and music, just I hadn’t connected it so much with Google previously. Furthermore, this year Art talks were mainly about the collaboration between artists and AI or ML, which is another topic close to my interests. I found it actually quite curious to see what are the artists’ learnings about the collaboration with AI. I think those could be an interesting base for improving human-AI work in other industries. Lastly, a nice message repeated throughout multiple talks was that AI helps you open new horizons as a tool and it does not replace you neither your work.
Security and privacy
must admit - learning about those topics wasn’t in my interest when going to the conference, but you just couldn’t avoid those words at most of the speeches. Yes, security and privacy were heavily mentioned topics during the conference, which nicely contrasted with a plane flying with a protest sign: “Google Control is Not Privacy #SaveLocalNews” banner :). But let’s talk examples: Google is now providing options to reset the data that are stored by its applications; making Android phones smart enough to recognize “hey Google” without sending every word you say to the cloud, or adding a physical switch to the new Google Nest that cuts off the camera and mic. And it’s left only to us to believe it… or test it ;)).
All in all, I am happy and grateful that I could attend the conference. It was a brilliant to experience how such a conference (and Silicon Valley :)) looks like in real, and to learn and network with amazing people. If you miss some spark of excitement in my tone, well, I believe this article might explain it a bit. If you have any comment or questions, please contact me or leave the comment.
Recommended links aboutEthical AI:
Designing Human-Centered AI Products talk
https://ai.google/education/responsible-ai-practices, https://cloud.google.com/inclusive-ml/#fairness-in-ml-automl, https://jigsaw.google.com/projects/ -Google projects and guidelines
https://github.com/tensorflow/tcav - Tensorflow model for interpreting ML predictions on a high level (increasing transparency of ML).
Recommended talks from AIxArt:
Art and Technology Experiments with Google Arts and Culture - 2nd in the list of recommended videos from me - this one brings smile immediately when I open it, a lot of entertainment there :)
Making Art with Artificial Intelligence: Artists in Conversation - for insights and inspiration
Music and Machine Learning - where you can The Flaming Lips artist playing with AI fruit bowl and Yacht describing their experience with creating lyrics with a help of Machine Learning.
As an add-on: 360 view of one of the greatest moments of The Flaming Lips concert on the 2nd day of Google I/O.
Other talks that either I or my friends recommend:
On Creativity and Technology, with Legendary Animator Glen Keane
Other links you might find useful: