May 28, 2026, 2:37 PMTimnit Gebru was fired from Google in December 2020 for refusing to retract a research paper, and every single warning that paper made about large language models has now happened at a scale the industry spent 4 years trying to make people forget about.
#AIEthics #WeDontNeedNoAIEthics #WeAintGotNoStinkeenAIEthics
May 28, 2026, 2:45 PM@teledyn don't need ethics when there is money to be made. Same thing happened with oil companies when their own research showed they were causing damage to the environment.
May 29, 2026, 10:03 AM@teledyn
just for the record. This paper is no a scientific paper, mentions no dataset, neither methodology, is what's called "position paper". There is no science there, no math, no STEM whatsoever, no data, no methodology.
Is a paper, sure. "According with me...."
May 30, 2026, 7:11 PM@uriel perhaps @emilymbender could clarify? Pretty please?
Jun 2, 2026, 2:32 AM@teledyn @uriel@bbs.keinpfusch.net @emilymbender
The link to the original paper is at the bottom of the post.
I'll repaste it here https://dl.acm.org/doi/epdf/10.1145/3442188.3445922
@uriel @teledyn It was published at the 2021 ACM Conference on Fairness, Accountability, and Transparency. It's not a conference on AI models, it's a conference about accountability and transparency of computing systems.
"Science" is more than math and data and methodology. It includes many kinds of analysis, including reasoning from first principles and making inductive and deductive arguments, as well as presentation of new ideas.
At a conference on fairness, accountability and transparency, I would expect to see a paper containing analysis of potential bias and obfuscation as well as recommendations for avoiding them, and I would have no problem counting it as a scientific paper if it makes a clear, sound argument for why a given thing will produce biased results and enable avoidance of accountability. The fact that these specific outcomes were later independently found to occur simply reinforces the conclusion that the analysis was sound.
The criteria for what counts as a "scientific paper" are amorphous. Different research areas have different norms. For decades, robotics conference papers basically consisted of "Look ma! No hands!"
The real surprise to me is that a conference accepted a 14 page paper. In my field, the norm is six pages and you can pay extra for up to a total of eight.
Link to paper: https://dl.acm.org/doi/10.1145/3442188.3445922
@Robotistry Pay extra for more pages? Now that's a model I've never heard about before.
But indeed, every field has their conventions. I don't think I've ever attended conferences that accept 14 page papers pre-conference, but I've been to several that do that post-conference with only an abstract pre-conference.
What I don't think a lot of people realise is that in many fields, the text *is* the research, or at least a big part of it, not just a report of the research.
Jun 1, 2026, 6:00 AM@teledyn Great, damning article. I'm now even more glad I've gotten almost completely free of Google (and Microsoft, and never used the others).
@teledyn Timnit is on mastodon if anyone wants to follow: @timnitGebru
@RakowskiBartosz also done, I have always had doubts about AI in totality... (Sometimes has its uses) @teledyn @timnitGebru
@RakowskiBartosz @teledyn @timnitGebru Timnit Gebru for president!
@teledyn They knew. They knew and did it anyway. All for a fast buck at our expense 🤬
I had to think about precisely this when the video of former Google CEO Eric Schmidt being booed at a university speech was making the rounds.
He told the students some stuff about how they’d shape the direction of new technological changes like AI. To be “the voice in the room”, implication being that these young students could somehow steer any new technological changes to be moral, and good.
And I’m thinking, Eric, Google had a voice like that and Google made her leave your room.
Edit: I want to be precise that Eric wasn’t CEO when Google fired Timnit Gebru but he has no business selling is the illusion of a seat at the table when we see evidence to the contrary
@astroPug @teledyn Eric is the good guy right? https://m.youtube.com/watch?v=tlQ7EoJDTQY&t=332s
Ah. So the Eric I mean is Eric Schmidt, the CEO who held the speech and was deservedly booed. I feel like he conveniently left out what happens to people who do use their voice.
I quite like the gentleman who made this video and the points he made. I don’t know his name - I wanna say Casey but I forgot - but the YouTube name is Molly Rocket.
@teledyn This is so often the fate of truth-tellers, especially the female kind.
A good article for the non-techy to read. Many of us have deep doubts and suspicions but don't have the language to express them. #TimnitGebru #AI #AIEthics
@teledyn Thanks, didn't know this. Her story is also well documented on Wikipedia: https://en.wikipedia.org/wiki/Timnit_Gebru and even her Mastodon-account is there @timnitGebru
Jun 2, 2026, 6:56 AM@teledyn ugh I just went down a rabbithole because I'm avoiding doing what I'm supposed to do and apparently that text was originally posted by Guri Singh on X, but the account was suspended: https://x.com/heygurisingh/status/2059251382960734593
And the reason it was suspended... was undisclosed AI use on that post, according to this: https://piunikaweb.com/2026/05/28/x-suspend-accounts-ai-replies/
Which was based on Pangram considering it AI-generated. AI-detection is shoddy, but I can see why people suspected it. It's just a classic 2026 AI mess, really.
Jun 2, 2026, 7:01 AM@teledyn Anyway, I really need to actually read Timnit et al.'s paper because it's obviously important and relevant and I love what DAIR is doing. It's just funny/horrifying how much of an entangled AI mess the Internet is becoming, where some AI critique is itself AI-generated for cheap clout (because it is justifiably popular) and detected (or not) by more AI-based tools. Blergh.
Jun 2, 2026, 9:55 AM@confuseacat @teledyn who the hell knows. Could have just been an excuse to remove a post that goes against Musk's AI-accelerationist tendencies. 🤷
@stragu @confuseacat @teledyn now that wouldn't shock me at all.
@teledyn The energy use of 5 cars? So optimistic! :-)
Jun 2, 2026, 9:07 AM
@teledyn Is there a better source for this than Tumblr?
Jun 2, 2026, 3:11 PM@teledyn No success, unfortunately...
Jun 2, 2026, 4:42 PMJun 2, 2026, 4:45 PM@davep @teledyn Thanks. I was interested in a link to the summary included in the original post, but I guess that can't be found.
Anyway, the original paper is excellent. I had known the expression "stochastic parrot" but did not know its origins. I wish I had known about it before. This paper really does anticipate most of the conversations we have had about LLMs in recent years. Really excellent work.