Timnit Gebru

@timnitgebru.blacksky.app

34521 Followers 903 Following

Personal Account Founder: The Distributed AI Research Institute @dairinstitute.bsky.social. Author: The View from Somewhere, a memoir & manifesto arguing for a technological future that serves our communities (to be published by One Signal / Atria

Nan Ransohoff @nanransohoff New blog post: The third wave of American philanthropy Hundreds of billions of dollars in new philanthropic capital will soon become liquid. The OpenAI Foundation holds 26% of OpenAI, worth about $220B at today’s valuation. Anthropic’s seven co-founders have pledged to give away 80% of their wealth and have instituted the most aggressive donor matching program for employees in tech history. How much does this all add up to? And how meaningful is that in the context of philanthropy today? I was doing some simple napkin math to wrap my head around the scale of what’s coming, and radicalized myself in the process. I had dramatically underappreciated the scale of the philanthropic capital that’s about to become available and the corresponding gap in talent and organizations that will be needed to make the most of it. This piece aims to directionally sketch the scale of what’s coming, the gap in operational capacity needed to absorb it, and what we can do to fill it. (Link to full post in reply)
Elon Musk, Benjamin Netayahu, Max Tegmark and another white guy I can't recognize.
Elon Musk, Benjamin Netanyahu, Max Tegmark and another white guy I don't recognize.
Rights Con 26 Zambia The leading summit on human rights in the digital age Important update from the RightsCon team We are aware of a media announcement indicating RightsCon has been postponed by the Government of Zambia and understand the panic it must be causing for our participants, especially those traveling to Lusaka. We have not yet received formal communication from the government and have requested an urgent meeting with the involved Ministries. We are on the ground coordinating with our partners and hope to have more information today (Wednesday, April
Imagining Possible Futures. Portland. Saturday April 11, 1pm--3pm. 3639 N Mississippi Ave Portland, OR
Alexander Kustov @akoustov.bsky.social Prof. @emilymbender.bsky.social, I appreciate you sharing your opinion on my integrity, which you're free to do even if it's not informed. But you're tenured too, so it's ok to admit you weren't right. You can say LLMs were trash before but now they can actually be useful to folks who use them. Quote tweeting Emily M. Bender saying "Prof. @akoustov.bsky.social, the point of tenure is academic freedom, but you seem to be either announcing to the world that you aren't interested in doing research, that you have no respect for the work (or time) of other scholars, or both.
Bryan (they/them) ‪@chaosgreml.in‬ Listen y’all, when you have a person this out of control on your list of “affiliates,” and one of your core team wrote a book with her, I start to question your values, @dairinstitute.bsky.social ¯\_(ツ)_/¯ Also, ZERO dark-skinned Black folk on the core team. Hm. ‪Alexander Kustov‬ ‪@akoustov.bsky.social‬ · 8h Quote tweeting Alexander Kustov: Your periodic reminder this all started when Dr. Bender, unprompted, decided to single me out as someone didn't deserve tenure for using LLMs. Her followers tagged my employer with calls to fire me and sent threats to me and my family. Everything since has been a response to that.
Flow chart showing OpenAI, Google Meta, Microsoft, Amazon, Perplexity, ProRate.ai, IBM, Mistral, Anthropic, Nvidia, in that order with the biggest deals, with OpenAI having 21 deals, Google 17, Meta 8, Microsoft 4, Amazon 3.
Nine, mostly white dudes on stage in a manel.
SCOOP: Anthropic was among the AI companies that submitted a proposal earlier this year to compete in a $100 million Pentagon prize challenge to produce technology for voice-controlled, autonomous drone swarming, acc to people familiar w/ matter.
There's a war where Israeli and Iranian drones fly side by side Deadly enemies are both supplying Ethiopia’s military – and civilians are in the firing line
We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax. These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges with Claude, extracting its capabilities to train and improve their own models.
Supporters Our work is supported by a variety of foundations, charities, and individuals who share our commitment to high-quality journalism about AI (grouped by lifetime giving): ​ $1M+ Coefficient Giving (formerly Open Philanthropy) (2023, 2024, 2025) Survival and Flourishing Fund (2024, 2025) $100k – $1M The Casey & Family Foundation (2025) EA Infrastructure Fund (2023) Future of Life Institute (2024) $10k – $100k ACX Grants (2024) AI Safety Tactical Opportunities Fund (2024) Cullen O'Keefe (2025) Hazel Browne (2024) Newman Family Charitable Fund​ (2025) Robert and Virginia Shiller Foundation (2023, 2024) ​​​ We have also received donations of less than $10,000 from a variety of generous individual donors. ​​ Our donors have no editorial control over the work of Tarbell, our fellows, or our grantees. Tarbell does not accept anonymous donations greater than $10,000. For details, see the Donor relations section of our ethics and standards policies.
Oh dear. Well I am familiar with it and consider it to be terrible scholarship I'm afraid Emile. But yes, it's hyperbolic, and demonstrates—in my view!—a very poor grasp of liberal political philosophy. But, as with Holly on another thread, I doubt I'll convince you of much on here
I'm not sure if it's worth adding this, but in case anyone's wondering, @AmandaAskell is a brilliant philosopher doing by far the most important philosophical research anyone in the world is doing right now (in terms of both impact and philosophical interest). I've had the privilege of seeing her at many stages along this journey, from a chance meeting in the hallways of Coombs (the building where ANU Philosophy used to be) back in about maybe 2015, to a visit with her at OpenAI's offices on the day, more or less, that she and others decided to leave OpenAI to form Anthropic, to a number of workshops and other things since, where she showed an incredible generosity of spirit to the folks both in philosophy and in the FAccT community who most needed to learn from her work.
(0:00) There's a particular reference Epstein makes to Africa, and he says, (0:04) I do a lot of work in Africa, and he refers to Africa as fertile ground for experimentation. (0:11) And so it turns out that Epstein was a major funder of this weird artificial intelligence (0:17) lab being created in Ethiopia that's tied to the scientist who's overtly transhumanist (0:24) that he funded extensively named Benjamin Goertzel. Goertzel is, for example, (0:29) head of Humanity Plus, formerly the World Transhumanist Association, (0:34) which is an entity that Epstein funded in addition to funding Goertzel's own work. (0:38) And Goertzel is also the chief scientist at Hanson Robotics, which has produced (0:43) these human-like robots, the best well-known probably being Sophia, the robot that was, (0:49) for example, granted citizenship by Mohammed bin Salman at a time when one of his close advisors (0:53) was Jeffrey Epstein. Again, very odd activity there. (0:58) So as far as what was going on in Ethiopia, at the time that Epstein was investing in this AI lab (1:06) and backing Goertzel, Goertzel was teaming up with this blockchain company called Cardano (1:13) that's pretty well-known in the cryptocurrency space, and they were specifically seeking to (1:20) offer digital ID—well, it's already happened. They gave a digital ID essentially to 5 million (1:27) Ethiopian schoolchildren with the intention of expanding that, and the whole intention of this (1:32) is to harvest data from these children. They explicitly say that. And even in other documents, (1:39) Goertzel essentially says that this data is being used to feed the AI that powers robots like Sophia (1:47) at Hanson Robotics and stuff.
In this way, TESCREALists have been able to divert resources toward trying to build AGI and stopping their version of an apocalypse in the far future, while dissuading the public from scrutinizing the actual harms that they cause in their attempts to build AGI. As another example, Max Tegmark, cofounder of the Future of Life Institute along with Jaan Tallinn, delivered a talk at the 2017 Effective Altruism conference (EA Global) in which he argued that “if we don’t improve our technology, we are doomed ... but with tech, life can flourish for billions of years” [122]. But in 2023 the Future of Life Institute circulated a widely publicized petition signed by Tegmark and many of those responsible for the AGI race, including Musk and Altman, to “pause giant AI experiments” to stop “nonhuman minds that might eventually outnumber, outsmart, obsolete and replace us” [123]. Appearing on Democracy Now!, when asked about the present day dangers like biometric surveillance warned by researchers and activists like Tawnana Petty, Tegmark stated: “Extinction is not something in the very distant future ... And once we’re all extinct, you know, all these other issues cease to even matter” (Bengio, et al., 2023). Similarly, when Geoffery Hinton, who also signed the petition, was asked by Rolling Stone about the issues raised by Timnit Gebru who was fired by Google after writing a paper on the dangers of large language models, he answered: “I believe that the possibility that digital intelligence will become much smarter than humans and will replace us as the apex intelligence is a more serious threat to humanity than bias and discrimination” (O’Neil, 2023).
9 almost all white men, including Bostrom, billionaires Musk and Tallin, discussing "AI" and "existential risks."
Board of directors "Virginia Blanton" Director of Operations at Anthropic.
This report taxonomizes and assesses these nuclear “fast-tracking” initiatives. We examine their feasibility and their impact on nuclear safety, security, and safeguards3—and more largely society’s potential exposure to radiation levels—across three primary categories: Policy initiatives seeking to lower regulatory practices including long-established nuclear-safety and risk-analysis approaches, safety culture, acceptable risks, and thresholds in order to reduce timescales for the construction of civil and defense nuclear facilities The use of generative AI to expedite regulatory processes such as nuclear licensing and commissioning for both civil and defense nuclear facilities The promotion of advanced and new nuclear technologies that are contingent on novel or unmaterialized approaches and infeasible timescales
A screenshot of the "Data for Change" section of the DAIR website (i.e. air-institute.org/projects and Data for Change, showing the Data Workers' Inquiry Project, Surveillance Watch and Impacts of Spatial Apartheid. The website is colorful, with orange and cream backgrounds.
From ICE Abductions to Genocide in Gaza, Languge Workers Resist. A Language Justice Forum with Respond Crisis Translatin. bit.ly/rct-rsvp. Oct 30, 2025. 5:30pm --SF Bay Area, CA
Sam Altman (@sama) We made ChatGPT pretty restrictive to make sure we were being careful with mental health issues. We realize this made it less useful/enjoyable to many users who had no mental health problems, but given the seriousness of the issue we wanted to get this right. Now that we have been able to mitigate the serious mental health issues and have new tools, we are going to be able to safely relax the restrictions in most cases. In a few weeks, we plan to put out a new version of ChatGPT that allows people to have a personality that behaves more like what people liked about 4o (we hope it will be better!). If you want your ChatGPT to respond in a very human-like way, or use a ton of emoji, or act like a friend, ChatGPT should do it (but only if you want it, not because we are usage-maxxing). In December, as we roll out age-gating more fully and as part of our “treat adult users like adults” principle, we will allow even more, like erotica for verified adults.
@BarackObama After two years of unimaginable loss and suffering for Israeli families and the people of Gaza, we should all be encouraged and relieved that an end to the conflict is within sight; that those hostages still being held will be reunited with their families; and that vital aid can start reaching those inside Gaza whose lives have been shattered. More than that, though, it now falls on Israelis and Palestinians, with the support of the U.S. and the entire world community, to begin the hard task of rebuilding Gaza – and to commit to a process that, by recognizing the common humanity and basic rights of both peoples, can achieve a lasting peace.
if you're writing a sentence that sounds like eugenics but you go "oh that's fine to say because it's not a real person" (whatever that means) you may want to consider what made you okay with saying that.
This broad (@timnitGebru) is so bad for black women, women in general & women from Tigray she claims to speak for. She's always lying, playing victim (while she is engaged in a narrative that *creates* victims) - & often does it citing ethnicity, gender or race. (throwing up emoji) #nomore
Report from Ethiopia, The Untold Story with Eugene Puryear & Hermela Aregawi, organized by The People's Forum an dBreakthrough News. Co-Sponsored by New Africa Institute and ANSWER coalition.
The following organizations are convening the confernece: Palestinian Youth Movement (PYM) US Palestinian Community Network The People's Forum (TPF) Al-Adwa The Palestinian Right to Return Coaltion. The People's Forum is circle.d
"Report from Ethiopia" The Untold Story w Eugene Puryear & Hermela Aregawi. Co Sponsored by New Africa institute and ANSWER Coaltion. Also has logos of The People's Forum, Breakthrough News.
Screenshot of Karen's headshot in her TIME/100 entry. Karen is a Chinese American woman with dark hair that is a little longer than her neck.
A headshot of Mila smiling and looking at the screen.
A black and white art showing a Black woman with an afro strapped on a chair and what looks like ghosts haunting her and surrounding her.
1. COMMUNITY: Reducing the distance between researchers and community collaborators Our research centers lived experiences from communities which are not typically represented in AI research. We strongly believe in a bottom-up approach, supporting and prioritizing research and researchers that emerge from the communities that we center in our work and their specific needs. Our approach comprises three key aspects. First, by fostering a distributed team, we reduce the distance between researcher and community collaborators––both geographically and in terms of our researchers’ contextual knowledge of and commitment to specific community needs. Second, we refuse to treat any individual as a tokenized stand-in for any community. Instead, we understand community as a network of shared trust, accountability, similar beliefs, bounds of solidarity, and shared resources. Finally, we realize that there is a long history of researchers co-opting research and knowledge authored by specific communities towards their own end. We work against that impulse by interrogating our positionality––understood as the social and political contexts that shape our individual outlook on the world––and knowing that this is an ongoing task. We aim to be continually reflexive about the negotiations that take place for the research to happen and the power differentials between researcher and community collaborators.
Screenshot of the DAIR research philosophy pdf with DAIR logo and art in the form of various shapes, mostly half circles. The background is orange, along with some yellow.
Mila's headshot with TIME 100/AI text.
Gary Marcus "I used to think (or at least hope) that @xriskology was wrong about this, that nobody could seriously want to replace humans with machines. But that was before @elonMusk switched from Team Human to Team "I Would Like to be Around to Watch." Quote tweeting Emile P. Torres's article on Digital Eugenics and the Extinction of Humanity.
Teaching staff and students under a big tree taking a group photo for JamCoders. Most people in the photo are Black, and the class is half girls and half boys. Teaching assistants are mostly towards the front and are a mix of Black, white and Asian men and women.
What are all the different things bucketed under “AI”? What questions should you ask when someone tells you to use “AI”? What are things that are useful under the tasks bucketed under “AI”? What is “hallucination” and why should you care about it? What is “reasoning” and why are claims of “reasoning models” bogus? How do detectors of “AI written content” work and why shouldn’t you rely on them? What are the energy costs of “AI”? What is “agentic AI” and why should you avoid it? What are alternatives to LLMs for the useful things bucketed under “AI”? What legalities should you worry about when you use ChatGPT and similar products (e.g. copyright, making stuff up, plagiarism) What is the real monetary cost of building ChatGPT and its likes, and why is it unsustainable (the prices will be hiked up like Uber, investors are currently subsidizing it, you’ll be paying more when they stop doing that). What security issues are you introducing when you use ChatGPT? What issues are you introducing when you generate code with Copilot and its likes?
Sam Altman & @sama btw, i think we will have these before this decade is out. not deployed at scale yet, but it will be clear that both are going to happen. Sam Altman& @sama • May 1, 2022 the future can be so good that it's hard for any of us to imagine. my basic take on this is that we will have 'unlimited' intelligence and energy, and all that will unlock. but the risks grow with the upside.
Friday April 17/2025 Public Notice Regarding Police raid on Addis Standard Detentions and mass equipment confiscations On Thursday April 17/2025, a coordinated police raid targeting a private house of a senior member and the office of Addis Standard, the tri-lingual publication by JAKENN Publishing PLC, has resulted in detention of senior staff members and confiscations by the police of several electronics devises belonging to our employees and the company. Details In the early hours of Thursday April 17/2025, a senior office of our IT department was forcibly taken from home at 5:00 AM by police forces. Our staff was physically mishandled during the incident; the security forces confiscated the following personal and work-related electronics items, 1 work laptop 2 mobile phones (one belonging to the spouse) 1 lash disk, 1 external hard disk. The staff member was also forcibly filmed by a police officer during the raid and was forced to sign on a paper admitting that they have willingly surrendered the equipment to the police officers. By midday, the senior staff was released, but all electronic devises remain under the control of the police. Office raid and mass equipment confiscations Later the same morning, at approximately 10:30 AM, six security personnel - who identified themselves as members of the Federal Police, arrived at our office in a white minivan. They seized multiple electronic devices and detained two of our senior colleagues from the newsroom and the HR department. Both were taken to the Federal Police Crime Investigation Unit near Mexico Square in the Capital Addis Abeba (towards Genet Hotel), where they endured an entire day of interrogation before being released in the evening.
"The war is with those you grew up with, your neighbors. If you can rid your forest of these thorns...victory will be yours!" Screenshot of facebook post in Amharic and its translation.
Yann LeCun At the end of 2017, the % of hate speech on FB taken down preemptively by AI was around 24%. Since early 2020, it's been above 94%. The rest is flagged by users and taken down "manually." How did the numbers get better? By using large language models!
brainstorm activity 1 Imagine it's 2050 and things have turned around dramatically for the better. What words would you use to describe this magnificently changed world?
Two Black women and a Black man floating in space with planets and other circular objects. The background is purple and their clothing is in yellow and brown.
Tweet by Gisele Navarro @ichbinGisele "“Collectively, the AI search engines provided incorrect answers to more than 60% of queries.” ~ @CJR Screenoshot of an image showing "Generative search tools were often confidently wrong in our study" showing baplots of different offerings like ChatGPT, Perplexity and Deepseek, and showing the percentage of wrong answers in red and right ones in green.
Chuck Schumer's post on Twitter. I abhor many of the opinions and policies that Mahmoud Khalil holds and supports, and have made my criticism of the antisemitic actions at Columbia loudly known. Mr. Khalil is also legal permanent resident here, and his wife, who is 8-months pregnant, is an American citizen. While he may well be in violation of various campus rules regarding how the protests were conducted last year, that is a matter for the university to pursue, and I have encouraged them to be much more robust in how they combat antisemitism and maintain a harassment-free campus that protects the safety and security of Jewish and other students. The Trump administration’s DHS must articulate any criminal charges or facts that would justify his detention or the initiation of deportation proceedings against him. If the administration cannot prove he has violated any criminal law to justify taking this severe action and is doing it for the opinions he has expressed, then that is wrong, they are violating the First Amendment protections we all enjoy and should drop their wrongheaded action.
Tesla stock taking a nose dive from $262.67 at previous close to $222.15 at the end of today.
Takeover Tesla Saturday March 8 at 12:00 4180 El Camino Real Palo Alto, CA94306 #TeslaTakeover Flyer screenshot
The letter’s signers include Timnit Gebru, AI ethics researcher and founder of DAIR (The Distributed AI Research Institute); Alex Hanna, director of research at DAIR; Abeba Birhane, senior fellow of trustworthy AI at the Mozilla Foundation; Emily Bender, professor of linguistics at the University of Washington; and Sarah Myers West, managing director of the AI Now Institute.
Map showing drone warfare casualties in Ethiopia in the Tigray, Amhara, Oromia and Somali regions.
In this video, Emmanuel Macron stated “The challenge of Africa, it is totally different, it is much deeper, it is civilizational, today. What are the problems in Africa? Failed states, the complex democratic transitions, demographic transitions, which is one of the main challenges facing Africa.” He then went on to say "a successful demographic transition when countries still have seven to eight children per woman — you can decide to spend billions of euros, you will not stabilize anything."
After investing in some of the most widely known companies working on AGI, Musk has now founded another startup, xAI, focused on the topic [90]. One of the company’s advisors is Dan Hendrycks, the executive and research director of the Center for AI Safety, which was awarded a grant of US$5,160,000 from Open Philanthropy [91]. A post coauthored by Hendrycks published on the Effective Altruism Forum, stated that he “was advised ... to get into AI to reduce [existential risk], and so settled on this rather than proprietary trading for earning to give” [92].
Statement from the listed authors of Stochastic Parrots on the “AI pause” letter Timnit Gebru (DAIR), Emily M. Bender (University of Washington), Angelina McMillan-Major (University of Washington), Margaret Mitchell (Hugging Face) March 31, 2023 Tl;dr: The harms from so-called AI are real and present and follow from the acts of people and corporations deploying automated systems. Regulatory efforts should focus on transparency, accountability and preventing exploitative labor practices.
BigTech companies are throwing money at artificial general intelligence (AGI), with CEOs telling us that it will come, bringing with it utopia, any day now (for the measly price of SEVEN TRILLION DOLLARS according to Sam Altman). How did AGI come to be a dominant goal in AI? What motivates the drive to attempt to build it? This page is dedicated to answering this question, starting with our paper, the TESCREAL Bundle: Eugenics and the Promise of Utopia through Artificial General Intelligence, which you can read on First Monday. Image is Clarote & AI4Media / Better Images of AI / Power/Profit / CC-BY 4.0
GPT-4.5 is ready! good news: it is the first model that feels like talking to a thoughtful person to me. i have had several moments where i've sat back in my chair and been astonished at getting actually good advice from an AI. bad news: it is a giant, expensive model. we really wanted to launch it to plus and pro at the same time, but we've been growing a lot and are out of GPUs. we will add tens of thousands of GPUs next week and roll it out to the plus tier then. (hundreds of thousands coming soon, and i'm pretty sure y'all will use every one we can rack up.) this isn't how we want to operate, but it's hard to perfectly predict growth surges that lead to GPU shortages. a heads up: this isn’t a reasoning model and won’t crush benchmarks. it’s a different kind of intelligence and there’s a magic to it i haven’t felt before. really excited for people to try it!
Screenshot of the panelists at the data labelers association launch event.