A major blowup at at Google led to the firing or resignation (depending on who you ask) of Timnit Gebru and accusations that the company's social-justice efforts are performative at best. But behind the scenes, some Googlers are telling a different, slightly more complicated story. Who's right? Katie and Jesse bring on Jon Stokes, an Ars Technica founder who is much better connected in the tech world than they are, to help explain this tangled controversy.
(Correction: In the original version of this episode we posted for our patrons, our guest, Jon Stokes, wrongly described Timnit Gebru's academic background as being in sociology, with her having also done some work in computer science. That should have been flipped: She has a PhD from the Stanford Artificial Intelligence Laboratory, but has done plenty of work that intersects with sociology. Jon also said she was educated at M.I.T., but all her degrees are from Stanford. Whenever an error makes it into the final show, that is 100% the fault of hosts, and we apologize. We've snipped out the audio containing the errors and added a link to Gebru's Stanford page to the top of the show notes.)
(More corrections, 12/18/2020: In the original versions of this pocast, both free and patrons-only, Jon also said at approximately 20:40 that Gebru had been at Google since 2016, though he hedged this with an "I think.". She had actually only been at Google since 2019 [YIKES -- actually 2018. Listen to the updated episode for a correction of the correction]. At about 56:30 [free version timestamp], he laid out a storyline, which he presented as informed speculation, in which Jeffrey Dean, head of Google AI, had promised Gebru and some of her allies in 2016 that they could do ethical work. Then, in this telling, Google pivoted to AI, and as a result Gebru and her colleagues' work was seen as a liability and came under more scrutiny. But that storyline doesn’t quite work, as Jon presented it, given that Gebru only arrived at Google in 2019, well after Sundar Pichai announced the company would be “AI first” in 2016. We've snipped out the audio containing these errors and will be discussing all of this episode's errors on the free episode of the podcast set to go live 12/21/2020, and earlier for patrons.)
Show notes/Links:
Timnit Gebru's background: https://ai.stanford.edu/~tgebru/
Jon Stokes: https://twitter.com/jonst0kes
His site, The Prepared: http://www.theprepared.com
The New York Times: Google Chief Apologizes for A.I. Researcher’s Dismissal - https://www.nytimes.com/2020/12/09/technology/timnit-gebru-google-pichai.html
The New York Times: Google Researcher Says She Was Fired Over Paper Highlighting Bias in A.I. - https://www.nytimes.com/2020/12/03/technology/google-researcher-timnit-gebru.html
Wired: Behind the Paper That Led to a Google Researcher’s Firing - https://www.wired.com/story/behind-paper-led-google-researchers-firing/
Platformer: The withering email that got an ethical AI researcher fired at Google - https://www.platformer.news/p/the-withering-email-that-got-an-ethical
MIT Technology Review: We read the paper that forced Timnit Gebru out of Google. Here’s what it says. - https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru/
Twitter: Yann LeCun on the data problem - https://twitter.com/ylecun/status/1274782757907030016
Twitter: Gebru disagrees - https://twitter.com/timnitGebru/status/1274808654227619840
(Note: As of 10:45 a.m. on Saturday I'm having trouble finding the throwaway Reddit posts we mention, but when I do I'll add those as well. Update: Thanks, John, for posting this in the comments: https://www.reddit.com/r/MachineLearning/comments/k77sxz/d_timnit_gebru_and_google_megathread/gepq3u8/?utm_source=reddit&utm_medium=web2x&context=3)
Share this post