#54: Google AI's ethics crisis, an adversarial attack on deepfake detectors, and Stanford's OGNet climate project

Hey everyone, welcome to Dynamically Typed #54! Today’s main story is about an ethics conflict at Google AI; I’ve done my best to summarize what happened over in the ML research section. Beyond that, I’ve got quick links across all categories: updates on adversarial deepfakes and Google Recorder for productized AI; a demo of speech recognition with overlapping voices for ML research; a new Stanford project to track oil and gas facilities, plus an Open Climate Fix job post for climate AI; and, finally, a robot arm that you can hand arbitrary objects to for cool things.

PS: I’m covering DeepMind’s AlphaFold breakthrough in the next edition of DT!

Productized Artificial Intelligence 🔌

• 🎠 Another episode in the saga of deepfakes, videos that make real people look like they’re saying or doing things they never said or did. In the fall of 2019, Facebook, Microsoft, and Google created datasets and challenges to automatically detect deepfakes (see DT #23); in October 2020, Microsoft then launched their Video Authenticator deepfake detection app (#48). Now, just a few months later, Neekhara et al. (2020) present an adversarial deepfake model that handily beats those detectors: “We perform our evaluations on the winning entries of the DeepFake Detection Challenge (DFDC) and demonstrate that they can be easily bypassed in a practical attack scenario.” And the carousel goes ‘round.
• 📱Google’s Recorder app for Android, which uses on-device AI to transcribe recordings (see DT #25, #31), now has a new ML-powered feature: Smart Scrolling. The feature “automatically marks important sections in the transcript, chooses the most representative keywords from each section, and then surfaces those keywords on the vertical scrollbar, like chapter headings.” This all happens on-device. How long until it also writes concise summaries of your hour-long recordings?

Machine Learning Research 🎛

Google AI is in the middle of an ethics crisis. Timnit Gebru, the AI ethics researcher behind Gender Shades (see DT #42), Datasheets for Datasets (#41), and much more, got pushed out of the company after a series of conflicts. Karen Hao for MIT Technology Review:

A series of tweets, leaked emails, and media articles showed that Gebru’s exit was the culmination of a conflict over [a critical] paper she co-authored. Jeff Dean, the head of Google AI, told colleagues in an internal email (which he has since put online) that the paper “didn’t meet our bar for publication” and that Gebru had said she would resign unless Google met a number of conditions, which it was unwilling to meet. Gebru tweeted that she had asked to negotiate “a last date” for her employment after she got back from vacation. She was cut off from her corporate email account before her return.

See Casey Newton’s coverage on his Platformer newsletter for both Gebru’s and Jeff Dean’s emails (and here for his extended statement). This story unfolded over the past week and is probably far from over, but from everything I’ve read so far — which is a __lot, hence this email hitting your inbox a bit later than usual — I think think Google management made the wrong call here. Their statement on the matter focuses on missing references in Gebru’s paper, but as Google Brain Montreal researcher Nicolas Le Roux points out:

… [The] easiest way to discriminate is to make stringent rules, then to decide when and for whom to enforce them. My submissions were always checked for disclosure of sensitive material, never for the quality of the literature review.

This is echoed by a top comment on HackerNews. From Gebru’s email, it sounds like frustrations had been building up for some time, and that the lack of transparency surrounding the internal rejection of this paper was simply the final straw. I think it would’ve been more productive for management to start a dialog with Gebru here — forcing a retraction, “accepting her resignation” immediately and then cutting off her email only served to escalate the situation.

Gebru’s research on the biases of large (compute-intensive) vision and language models is much harder to do without the resources of a large company like Google. This is a problem that academic ethics researchers often run into; OpenAI’s Jack Clark, who gave feedback on Gebru’s paper, has also pointed this out. I always found it admirable that Google AI, as a research organization, intellectually had the space for voices like Gebru’s to critically investigate these things. It’s a shame that it was not able to sustain an environment in which this could be fostered.

In the end, beside the ethical issues, I think Google’s handling of this situation was also a big strategic misstep. 1500 Googlers and 2100 others have signed an open letter supporting Gebru. Researchers from UC Berkeley and the University of Washington said this will have “a chilling effect” on the field. Apple and Twitter are publicly poaching Google’s AI ethics researchers. Even mainstream outlets like The Washington Post and The New York Times have picked up the story. In the week leading up to NeurIPS and the Black in AI workshop there, is this a better outcome for Google AI than letting an internal researcher submit a conference paper critical of large language models?

Quick ML research + resource links 🎛 (see all 70+ resources)

• 📲 Cool video demo of speech recognition for overlapping voices by Google AI’s Quan Wang: VoiceFilter-Lite lets users enroll their voices on their phones (“this is my phone, please remember my voice”) and is then very accurately able to filter out other peoples’ voices while transcribing text. This all happens on-device and has super low latencies.

Artificial Intelligence for the Climate Crisis 🌍

• 🛢 Researchers from Andrew Ng’s Stanford ML Group have “developed and deployed a deep learning model called OGNet to detect oil and gas infrastructure in aerial imagery.” They’ve used it to create an open dataset with 7,000+ images of nearly 150 oil refineries — including several facilities that were not yet included in existing public datasets — which they hope will make it easier to attribute satellite-detected methane emissions to their sources on the ground. The paper by Sheng & Irving et al. (2020) will be published at the 2020 NeurIPS workshop on Tackling Climate Change with Machine Learning. Research like this has the potential to become a key tool for climate policy makers and enforcers.
• 💼 Open Climate Fix, which works on solar PV mapping and nowcasting, among other things, is hiring an ML Research Engineer to “research cutting-edge ML to help mitigate climate change” and build out open-source services. Go apply! (#nospon)

Cool Things ✨

• 🦾 This is a bit out of the scope of what I usually cover on DT, but I was obsessed with robot arms during high school and this new NVIDIA paper by Yang et al. (2020) looks awesome. Their project, Reactive Human-to-Robot Handovers of Arbitrary Objects, does exactly what it says on the tin: it uses computer vision to let the robot arm grasp arbitrary objects presented by the user. This is a really difficult problem that’s key to building the kinds of robots we see in movies! The researchers posted a 3-minute demo video on YouTube, which is a fun watch.

Thanks for reading! As usual, you can let me know what you thought of today’s issue using the buttons below or by replying to this email. If you’re new here, check out the Dynamically Typed archives or subscribe below to get a new issues in your inbox every second Sunday.

If you enjoyed this issue of Dynamically Typed, why not forward it to a friend? It’s by far the best thing you can do to help me grow this newsletter. ⚓️