# Timnit Gebru [Frontier Minds profile](https://frontierminds.ai/people/timnit-gebru/) Last verified: 2026-09-19 Positions and narratives are editorial summaries. Quotes are attributed quotations. Source dates and profile verification dates are distinct. Missing positions mean not recorded, not agreement or neutrality. Founder and Executive Director · DAIR Institute San Francisco Bay Area, USA Electrical engineer and computer scientist who co-wrote Gender Shades, Datasheets for Datasets and Stochastic Parrots, left Google in a disputed exit in December 2020, and runs the Distributed AI Research Institute. ## Overview ### Addis Ababa, Ireland and Somerville Gebru grew up in Addis Ababa with her mother, an economist. Her father, an electrical engineer with a PhD, died when she was small, and her two older sisters, engineers like him, had already moved to the United States. The family was Eritrean, and when war broke out between Ethiopia and Eritrea in May 1998, relatives began to be deported to Eritrea and conscripted. Gebru was 15. Her mother held a US visa; Gebru's own application was refused, so she went to Ireland to join a sister who was there for work, while her mother went to America alone. She has said she called her mother from Ireland and begged to be sent back to Ethiopia. The following year she was admitted to the United States as a refugee and joined her mother in Somerville, Massachusetts, where she enrolled in the public high school and found that some teachers would not accept that an African refugee could top the class in math and physics. The Carnegie Corporation, which named her a Great Immigrant in 2023, records that she was 16 when she sought asylum. She entered Stanford in 2001 and chose the family subject, electrical engineering. A junior-year course project, an experimental electronic piano key, won her an internship at Apple making audio circuitry; the next year she joined the company full time while finishing her degrees. Her manager, Niel Warren, later told Wired that "as an electrical engineer she was fearless": when he needed someone to investigate whether delta-sigma modulators would work in the iPhone, she volunteered. Her own summary of the Apple years is that she designed circuits and signal-processing algorithms for products including the first iPad. She also spent, in her words, "an obligatory year as an entrepreneur," and in 2008 dropped a class to canvass for Barack Obama in Nevada and Colorado. ### Counting cars, then counting colleagues Gebru returned to Stanford for a PhD in Fei-Fei Li's lab, supported by an NSF Graduate Research Fellowship and a Stanford DARE fellowship. Two of Li's students had scraped 50 million Google Street View images to train a network to recognize cars. Gebru's project asked what else the cars could reveal. Using online contractors and car experts recruited on Craigslist, her team labeled the make and model of 70,000 vehicles, trained detectors on them, and ran the detectors over images from 200 US cities, cataloguing 22 million cars, roughly 8 percent of all the automobiles in the country. Correlated with census and election data, the counts predicted income, race, education and votes: if sedans outnumbered pickup trucks on a 15-minute drive through a city, the paper reported, the city was likely to vote Democratic (88 percent), and otherwise Republican (82 percent). The work appeared in the Proceedings of the National Academy of Sciences in 2017. It was also, she came to think, a surveillance tool, and the thesis she finished that year closed with a pivot toward studying bias instead. Two conferences changed her direction. At NIPS 2015 in Montreal, strangers in Google Research T-shirts treated a Black woman's presence as a photo opportunity; one grabbed her for a hug, another kissed her cheek. At NIPS 2016 in Barcelona she counted the Black attendees she met: six, herself included, all people she already knew. She posted afterwards that she was not worried about machines taking over the world but about "groupthink, insularity, and arrogance in the AI community," and the email list she had started for Black researchers became Black in AI, co-founded with Rediet Abebe in 2017. Its first workshop, at NIPS 2017 in Long Beach, is where Jeff Dean and Samy Bengio of Google Brain first suggested she join them. In the summer of 2017 she took a postdoc in Microsoft Research's FATE group in New York, and with Joy Buolamwini published Gender Shades in February 2018. ### Google and the Stochastic Parrots dispute Gebru joined Google in 2018 as a research scientist and, with Margaret Mitchell, co-lead of the Ethical AI team. Nature later reported that Black women made up 1.6 percent of Google's researchers at the time. Her Google years produced Model Cards for Model Reporting (2019), the internal auditing framework Closing the AI Accountability Gap (2020), and, in April 2019, her signature on an open letter from 26 researchers, Yoshua Bengio among them, asking Amazon to stop selling facial recognition to police. In late 2020 she, Mitchell, Emily Bender, Angelina McMillan-Major and two other Google colleagues submitted "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?" to the FAccT conference. Jeff Dean, Google's head of research, had encouraged her to think about the downsides of large language models, and the paper had cleared Google's usual publication review. Then, by her account, an executive named Megan Kacholia ordered her in a last-minute video call to retract it or remove the Google authors' names, with a deadline the day after Thanksgiving. The reviewers' objections came in a document that her manager, Samy Bengio, was told to read aloud to her rather than send. Gebru asked Google, in writing, to identify who had reviewed the paper and how, and to set up a more transparent process; if not, she wrote, she would work out a last date. A second email, to a list for women at Google Brain, accused the company of "silencing marginalized voices" and asked, "Have you ever heard of someone getting 'feedback' on a paper through a privileged and confidential document to HR?" The next night, at an Airbnb in Austin, a direct report messaged her: "You resigned??" Kacholia's email to her personal inbox said, "We cannot agree as you are requesting," and, "The end of your employment should happen faster than your email reflects." Google's account differed on the process and the ending. In a memo to Google Research published by Platformer, Dean wrote that the company requires two weeks for this kind of review, that "this particular paper was only shared with a day's notice before its deadline," that a cross-functional team found it "didn't meet our bar for publication" because it ignored recent work on making models more efficient and less biased, and, "We accept and respect her decision to resign from Google." Gebru maintains she was fired; Google maintains she resigned. On 9 December 2020 Sundar Pichai told employees that her departure "seeded doubts and led some in our community to question their place at Google," said "I want to say how sorry I am for that," and ordered a review of what had happened. By December, according to Nature, almost 7,000 researchers and engineers, more than 2,600 of them at Google, had signed a petition demanding an overhaul of the company's research integrity, and nine Democratic members of Congress wrote to Pichai. Mitchell, who had searched her corporate email for evidence of discrimination against Gebru, was suspended in January and fired on 19 February 2021. Dean announced that progress on workforce diversity would count in senior executives' performance reviews and that researchers would get firmer guidance on Google's research priorities. Samy Bengio left for Apple. The paper itself was published at FAccT in March 2021 and, as of September 2026, is cited more than 17,000 times. ### DAIR On 2 December 2021, a year to the day after her exit, Gebru launched the Distributed Artificial Intelligence Research Institute with $3.7 million from the Ford Foundation, the MacArthur Foundation, the Kapor Center and the Open Society Foundations, under the fiscal sponsorship of Code for Science and Society. The launch statement described "an independent, community-rooted institute set to counter Big Tech's pervasive influence on the research, development and deployment of AI." The first research fellow, Raesetje Sefala, used satellite imagery to trace the persistence of apartheid-era spatial planning in South African townships; Safiya Noble joined the advisory board; and in February 2022 the sociologist Alex Hanna left Google to become director of research. Gebru told IEEE Spectrum that DAIR would deliberately "put out less work, so each work will take more money," and that the open question was how to compensate people whose knowledge the institute drew on. DAIR is headquartered in Oakland, California, with staff across Africa, Europe and North America. Its largest project, the Data Workers' Inquiry, led by research lead Milagros Miceli with the Weizenbaum Institute, has 19 data workers in nine countries researching their own workplaces, and DAIR helped its Kenyan participants form the Data Labelers Association. Other programs include Surveillance Watch, an open repository documenting the spyware industry; the Huniki Federation, which supports small organizations building task-specific language models rather than one model for everything; Sovereign Language Technologies for East African languages; and, added in 2026, a research program on how refugees and migrants are affected by surveillance technology. The institute also produces the podcast Mystery AI Hype Theater 3000 with Emily Bender and Hanna. ### Against the machine-god When the Future of Life Institute called in March 2023 for a six-month pause on training systems more powerful than GPT-4, Gebru, Bender and McMillan-Major answered within days. The letter, they wrote, steered attention toward "the risks of imagined 'powerful digital minds'" and "addresses none of the ongoing harms from these systems," listing worker exploitation, mass data appropriation, synthetic media and the concentration of power. Hypothetical extinction risks, they argued, were "the focus of a dangerous ideology called longtermism." In May 2023, when Geoffrey Hinton left Google to warn of existential risk and was asked on CNN why he had not spoken up for Gebru, he said her concerns "weren't as existentially serious." With the philosopher Emile P. Torres she then set out to name what she saw as the ideology. Their paper in First Monday in April 2024 grouped transhumanism, extropianism, singularitarianism, cosmism, rationalism, effective altruism and longtermism into the "TESCREAL bundle" and argued that the normative framework behind the race to AGI "is rooted in the Anglo-American eugenics tradition of the twentieth century." Its engineering argument is simpler than its intellectual history: a system with no specified task cannot be tested for safety the way a bridge or an aircraft can, so researchers should build scoped systems with testable safety protocols instead. ### The book and the record so far Gebru was named to Nature's 10 in December 2021, Fortune's list of the world's greatest leaders (24th) in 2021, the TIME 100 in 2022, where Safiya Noble's tribute called her "a truth teller," and the BBC's 100 Women and the Carnegie Corporation's Great Immigrants in 2023. In February 2025 the National Information Standards Organization gave her its Miles Conrad lifetime achievement award for the influence of datasheets and model cards on information practice; its announcement described her memoir under a working title, The View from Somewhere. By August 2026 the book had become Deep Unlearning: The Rise of AI and the Radicalization of a Tech Idealist, scheduled by Atria/One Signal for 16 February 2027, and she was touring it: a Scientific American interview in June 2026 in which she called American AI research "corporate-driven and sloppy," Democracy Now! in August, and a Wired interview in September in which she compared AI accountability to a bridge collapse and said the doom narrative was "meant to distract us." When Dario Amodei called on 12 September for the industry to "pace the frontier," she told CNBC that companies' claims of existential risk were overblown and that "There is no AI exemption in existing statutes." On 15 September 2026 SXSW announced her as a 2027 keynote speaker. Her Google Scholar profile lists 48,987 citations, an h-index of 32, and more than 13,000 new citations in 2026 alone. ## Timeline ### Leaves Ethiopia at 15 September 1998 After the Eritrean-Ethiopian War began in May 1998 and Eritrean relatives were deported, she was refused a US visa and went to Ireland; the next year she entered the US as a refugee and finished high school in Somerville, Massachusetts. ### Enters Stanford September 2001 Studied electrical engineering; a junior-year piano-key project led to an Apple internship in audio hardware and a full-time job the following year. ### Returns to Stanford for a PhD in Fei-Fei Li's lab September 2013 Began the Street View project on estimating neighborhood demographics from cars, funded by NSF and Stanford DARE fellowships. ### Counts six Black attendees at NIPS in Barcelona December 2016 The count, and a Facebook post about "groupthink, insularity, and arrogance in the AI community," led to the founding of Black in AI with Rediet Abebe in 2017. ### Joins Microsoft Research as a postdoc July 2017 Took a position in the FATE group in New York, having decided she wanted to help contain the technology's power rather than expand it. ### PhD, Stanford September 2017 Her thesis on computer vision for sociology closed with a turn toward studying algorithmic bias. ### First Black in AI workshop December 2017 Held at NIPS in Long Beach, where Jeff Dean and Samy Bengio of Google Brain first suggested she join them. ### Street View study in PNAS December 2017 The paper inferred income, race, education and votes in 200 US cities from 22 million cars. ### Gender Shades February 2018 With Joy Buolamwini, showed three commercial gender classifiers erred on darker-skinned women up to 34.7 percent of the time. ### Datasheets for Datasets first circulates March 2018 Posted to arXiv; the final version appeared in Communications of the ACM in December 2021. ### Joins Google September 2018 Research scientist and co-lead of the Ethical AI team with Margaret Mitchell. ### Model Cards for Model Reporting January 2019 Co-authored with Mitchell and colleagues at FAT\*; it became the template for model documentation on Hugging Face. ### The Amazon letter April 2019 Signed the open letter from 26 researchers asking Amazon to stop selling facial recognition to law enforcement. ### Exit from Google December 2020 Ordered to retract or remove Google names from Stochastic Parrots, she set conditions and was told her employment would end "faster than your email reflects"; she says she was fired, Google says she resigned. Pichai apologized to staff on 9 December. ### Margaret Mitchell fired February 2021 Her Ethical AI co-lead, who had searched her corporate email for evidence of discrimination against Gebru, was fired on 19 February. ### Stochastic Parrots published March 2021 The paper appeared at FAccT and became her most-cited work. ### DAIR founded December 2021 Launched a year to the day after her exit, with $3.7 million from four foundations. ### Nature's 10 December 2021 Nature named her one of the ten people who shaped science that year. ### Alex Hanna joins DAIR February 2022 The sociologist left Google to become DAIR's director of research. ### TIME 100 May 2022 Listed among TIME's 100 most influential people, with a tribute by Safiya Noble. ### Response to the AI pause letter March 2023 With Bender and McMillan-Major, argued the Future of Life Institute letter ignored present harms and rested on "a dangerous ideology called longtermism." ### Carnegie Great Immigrants honoree June 2023 The Carnegie Corporation of New York's citation records that she was 16 when she sought asylum. ### BBC 100 Women November 2023 Named to the BBC's list of 100 inspiring and influential women. ### The TESCREAL bundle April 2024 With Emile P. Torres in First Monday, traced the AGI project to the eugenics tradition and argued undefined systems cannot be tested for safety. ### Named 2025 Miles Conrad awardee December 2024 NISO's lifetime achievement award; she gave the Miles Conrad Lecture virtually at NISO Plus in Baltimore on 12 February 2025. ### Scientific American interview June 2026 Called American AI research "corporate-driven and sloppy." ### Deep Unlearning announced August 2026 The memoir went on preorder for publication on 16 February 2027. ### Wired interview September 2026 Told Wired the "machine-god narrative" is "meant to distract us." ### SXSW keynote announced September 2026 SXSW named her a 2027 keynote speaker. ## Key contributions ### [Gender Shades](https://proceedings.mlr.press/v81/buolamwini18a.html) With Joy Buolamwini, Gebru built a face dataset balanced by gender and by Fitzpatrick skin type, then ran three commercial gender classifiers over it. Darker-skinned women were misclassified up to 34.7 percent of the time; the worst error rate for lighter-skinned men was 0.8 percent. Published at the first FAT\* conference in February 2018, the paper made "intersectional" a benchmarking term, gave the campaign against police facial recognition its most-cited number, and framed the April 2019 letter to Amazon. It has more than 12,000 citations. ### [Datasheets for Datasets](https://arxiv.org/abs/1803.09010) Every electronic component ships with a datasheet listing its operating characteristics, test results and recommended uses. Gebru, who had spent years reading them at Apple, proposed in March 2018 that every training dataset should too: its motivation, composition, collection process, preprocessing, recommended and discouraged uses. The paper, with Jamie Morgenstern, Briana Vecchione, Jennifer Wortman Vaughan, Hanna Wallach, Hal Daume III and Kate Crawford, went through eight arXiv revisions before Communications of the ACM published it in December 2021, and it is cited about 5,500 times. ### [Model Cards for Model Reporting](https://arxiv.org/abs/1810.03993) Led by Margaret Mitchell with Gebru as senior author, the FAT\* 2019 paper proposed a short standard document for every trained model: intended uses, out-of-scope uses, and evaluation broken out by demographic, cultural and phenotypic group. Hugging Face's Hub documentation still cites "Mitchell, 2018" as the template for the README that accompanies every model it hosts. ### [On the Dangers of Stochastic Parrots](https://dl.acm.org/doi/10.1145/3442188.3445922) Written with Emily Bender, Angelina McMillan-Major and Margaret Mitchell and published at FAccT in March 2021, the paper sorted the risks of ever-larger language models into four bins: environmental and financial cost; uncurated web-scale data that encodes dominant viewpoints; the research opportunity cost of chasing scale; and text that is fluent enough to be mistaken for meaning. The phrase gave the field a name for that last failure. With 17,110 citations on Google Scholar, it is the most-cited paper in AI ethics. ### [Internal algorithmic auditing](https://arxiv.org/abs/2001.00973) Closing the AI Accountability Gap, led by Inioluwa Deborah Raji with Gebru, Mitchell and Google colleagues at FAT\* 2020, set out an end-to-end process (scoping, mapping, artifact collection, testing and reflection) for auditing AI systems inside the organizations that build them, with each stage producing documents that feed a final audit report. It is cited about 3,100 times and is the paper most often invoked when regulators describe pre-deployment audits. ### [Computer vision as social science, and its hazards](https://arxiv.org/abs/1702.06683) Her PhD work catalogued 22 million cars in 50 million Street View images from 200 US cities and used them to estimate income, race, education and voting patterns down to the precinct. The 2017 PNAS paper showed what large-scale image analysis could infer about people who had never been asked, and Gebru cites it as the work that turned her toward studying the harms of the methods she had used. ### [Black in AI and DAIR](https://www.dair-institute.org/) Black in AI, co-founded with Rediet Abebe in 2017, has granted more than $2 million to bring more than 400 practitioners to major conferences. DAIR, founded in December 2021, runs the Data Workers' Inquiry, Surveillance Watch and the Huniki Federation of community-built language models, and is Gebru's working answer to her own question of whether ethics research can survive inside the companies it studies. ### [The TESCREAL critique of AGI](https://firstmonday.org/ojs/index.php/fm/article/view/13636) With Emile P. Torres, Gebru named the cluster of ideologies she argues drives the race to artificial general intelligence and traced its lineage to twentieth-century eugenics. The April 2024 First Monday paper supplied the vocabulary that critics of effective altruism and longtermism now use, and its engineering claim, that an undefined system cannot be tested for safety, is the argument she returns to in every interview since. ## Signature ideas ### [Documentation is the first form of accountability](https://arxiv.org/abs/1803.09010) Gebru's most widely adopted idea came from her hardware years. Every chip Apple bought arrived with a datasheet describing what it was tested for and what it should not be used for, and nothing of the kind existed for the datasets and models that machine learning teams passed around. Datasheets for Datasets (first posted March 2018) and Model Cards for Model Reporting (2019) asked for that discipline: who collected the data and why, what is in it, what the model was evaluated on and for whom it fails. The proposal spread because it was cheap and legible to engineers; Hugging Face made a model card the README of every hosted model, and the audit framework she co-wrote in 2020 treats each stage's documents as the audit trail. The idea's limit is that a datasheet is voluntary and can be written badly. In June 2026 she said the basic practices she had asked for, "transparency, appropriate documentation," were still "just not being followed" across the field, which is her own evidence that documentation without independent enforcement does not change incentives. ### [Stochastic parrots, or fluency is not understanding](https://dl.acm.org/doi/10.1145/3442188.3445922) The 2021 paper's central claim is that a language model is a system for stitching together sequences of linguistic forms it has observed, without reference to meaning, and that the danger lies in how convincing the output is to the person reading it. Around that claim the paper arranged three others: training costs that fall on people who never use the products, web-scale data that over-represents whoever is loudest online, and research budgets diverted from smaller, better-understood approaches. The framing was contested from the start. Google's Jeff Dean said the paper ignored work on efficiency and bias mitigation; Dario Amodei and Sam Altman argue that scale has since bought scientific and economic gains that dwarf the costs, and Amodei's 2024 essay imagines powerful AI compressing decades of biological progress into years. Gebru's reply, in Wired in September 2026, is that OpenAI's claim in 2019 that GPT-2 was too dangerous to release shows the hype and the doom story are the same marketing, and that "you're asking the wrong questions." The phrase outlived the dispute: "stochastic parrot" is now the standard shorthand for the illusion-of-meaning problem. ### [AGI is an ideology with a lineage, not an engineering target](https://firstmonday.org/ojs/index.php/fm/article/view/13636) With Emile P. Torres, Gebru argued in 2024 that the drive to build a single general intelligence is inherited from a family of movements (transhumanism, extropianism, singularitarianism, cosmism, rationalism, effective altruism, longtermism) whose common ancestor is the Anglo-American eugenics of the early twentieth century, with its ranking of minds and its faith in engineering a better humanity. The historical claim is the controversial part; the engineering claim underneath it is that "undefined systems like 'AGI' cannot be appropriately tested for safety," because you cannot write a test plan for a product with no specified task. Yoshua Bengio's answer is that how a system reaches expert-level capability "does not change the existence of the risk," and that what matters is what it can do. Gebru's counter is that a field which cannot define the thing it is racing to build should stop calling the race science. The acronym is now common currency among critics of effective altruism, and among some of its defenders, who use it to describe what they are being accused of. ### [Community-rooted, slow research](https://www.dair-institute.org/projects/) DAIR is built on the premise that where the money comes from decides what questions get asked. Its founding rules were to take no corporate funding, to publish fewer papers and spend more on each, to treat communicating results to the public as part of the work, and to pay the people whose knowledge a project depends on. The Data Workers' Inquiry is the clearest case: instead of researchers interviewing data labelers, 19 labelers in nine countries were funded to investigate their own workplaces, and the Kenyan group went on to form the Data Labelers Association. The approach trades scale for standing. DAIR's output is small next to a university lab's, its projects are hard to replicate elsewhere, and its independence rests on a handful of foundations. Gebru's own framing in 2026 was that "the constraints breed innovation," pointing to small organizations that build their own compute clusters rather than rent from Google or Amazon. ### [Can a bridge decide to collapse?](https://www.wired.com/story/one-of-ais-fiercest-critics-says-all-the-doom-talk-is-meant-to-distract-us/) The line, from her forthcoming book, condenses her view of AI safety into a question of engineering liability. When a bridge falls, no one asks whether it was sentient; investigators ask who built it, who permitted it, and which tests were skipped. Talk of models "going rogue," she argues, moves the conversation from those questions to the machine's supposed inner life, which conveniently has no owner. On this view the things that are "really existential" are autonomous weapons already in use, the climate cost of data centers, and employers using AI as a reason to cut staff. Hinton and Bengio, who both signed the May 2023 statement that extinction from AI should be a global priority alongside pandemics and nuclear war, regard this as a false choice; Bengio has written that a democracy can discuss long-term and short-term harms at once, as it does with climate change. The disagreement is less about facts than about which harms deserve the scarce attention of regulators, and it has not narrowed since 2023. ## Notable works ### [Using Deep Learning and Google Street View to Estimate the Demographic Makeup of Neighborhoods Across the United States](https://arxiv.org/abs/1702.06683) paper · 2017 PNAS 114(50); the PhD study inferring neighborhood demographics from 22 million cars. ### [Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification](https://proceedings.mlr.press/v81/buolamwini18a.html) paper · 2018 With Joy Buolamwini at FAT\* 2018; the 34.7 percent versus 0.8 percent result. ### [Datasheets for Datasets](https://arxiv.org/abs/1803.09010) paper · 2021 First posted March 2018; published in Communications of the ACM 64(12) in December 2021. ### [Model Cards for Model Reporting](https://arxiv.org/abs/1810.03993) paper · 2019 With Margaret Mitchell and colleagues at FAT\* 2019; the template for model documentation on Hugging Face. ### [Closing the AI Accountability Gap: Defining an End-to-End Framework for Internal Algorithmic Auditing](https://arxiv.org/abs/2001.00973) paper · 2020 With Inioluwa Deborah Raji and others at FAT\* 2020. ### [On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?](https://dl.acm.org/doi/10.1145/3442188.3445922) paper · 2021 FAccT 2021; the paper at the center of her departure from Google and her most-cited work. ### [Announcing DAIR](https://dair-institute.org/press-release/) essay · 2021 The 2 December 2021 launch statement, with the line "AI needs to be brought back down to earth." ### [Timnit Gebru Is Building a Slow AI Movement](https://spectrum.ieee.org/timnit-gebru-dair-ai-ethics) interview · 2022 IEEE Spectrum, March 2022, with Alex Hanna, on why DAIR will publish less. ### [Statement from the listed authors of Stochastic Parrots on the 'AI pause' letter](https://www.dair-institute.org/blog/letter-statement-March2023/) essay · 2023 With Emily Bender and Angelina McMillan-Major, 31 March 2023. ### [The TESCREAL bundle: Eugenics and the promise of utopia through artificial general intelligence](https://firstmonday.org/ojs/index.php/fm/article/view/13636) paper · 2024 With Emile P. Torres, First Monday 29(4), April 2024. ### [Timnit Gebru on how to safeguard independent science for the AI age](https://www.scientificamerican.com/article/timnit-gebru/) interview · 2026 Scientific American, June 2026, on decoupling research funding from corporations and the military. ### [One of AI's Fiercest Critics Says All the Doom Talk Is 'Meant to Distract Us'](https://www.wired.com/story/one-of-ais-fiercest-critics-says-all-the-doom-talk-is-meant-to-distract-us/) interview · 2026 Wired, September 2026; the bridge-collapse argument and a preview of the book. ### [Deep Unlearning: The Rise of AI and the Radicalization of a Tech Idealist](https://books.apple.com/us/book/deep-unlearning/id6780517958) book · 2027 Memoir and manifesto from Atria/One Signal, due 16 February 2027; earlier announced under the working title The View from Somewhere. ## Where to start ### [What Really Happened When Google Ousted Timnit Gebru (Wired)](https://www.wired.com/story/google-timnit-gebru-ai-what-really-happened/) essay Tom Simonite's June 2021 reconstruction is the fullest single account of her life and of the Google dispute, with both sides' documents; about 45 minutes. ### [Gender Shades](https://proceedings.mlr.press/v81/buolamwini18a.html) paper Eleven pages, one method and one table; read it to see why "34.7 percent versus 0.8 percent" became the number that followed facial recognition into legislatures. Half an hour. ### [On the Dangers of Stochastic Parrots](https://dl.acm.org/doi/10.1145/3442188.3445922) paper The paper itself is calmer than the fight over it; the section on "illusions of meaning" is the origin of the phrase. About an hour. ### [Statement on the "AI pause" letter](https://www.dair-institute.org/blog/letter-statement-March2023/) essay Two pages, March 2023, and the clearest statement of the present-harms position in her own words. Ten minutes. ### [The TESCREAL bundle](https://firstmonday.org/ojs/index.php/fm/article/view/13636) paper Long and polemical, but it is where the argument that AGI is an ideology with a history is made in full; skim the genealogy, read the safety-testing section closely. Two hours. ### [One of AI's Fiercest Critics Says All the Doom Talk Is 'Meant to Distract Us' (Wired)](https://www.wired.com/story/one-of-ais-fiercest-critics-says-all-the-doom-talk-is-meant-to-distract-us/) interview The September 2026 interview is the current state of her thinking, with the bridge-collapse argument and a preview of Deep Unlearning. Fifteen minutes. ## Awards ### Fortune's World's 50 Greatest Leaders 2021 Ranked 24th, listed as co-founder of Black in AI. ### Nature's 10 2021 Named one of ten people who helped shape science in 2021, as "AI ethics leader." ### TIME 100 2022 Listed among the world's 100 most influential people, with a tribute by Safiya Noble. ### Carnegie Corporation of New York Great Immigrants honoree 2023 ### BBC 100 Women 2023 ### NISO Miles Conrad Award 2025 The National Information Standards Organization's lifetime achievement award; lecture delivered 12 February 2025. ## Education ### BS and MS in Electrical Engineering Stanford University ### PhD in Electrical Engineering (computer vision) 2017 · Stanford University ## Perspectives ### AI existential risk and longtermism Editorial summary: Gebru argues that extinction scenarios are the product of a longtermist ideology and that they pull attention and policy away from harms that are already measurable, from exploited data workers to autonomous weapons. > Those hypothetical risks are the focus of a dangerous ideology called longtermism that ignores the actual harms resulting from the deployment of AI systems today. Source: [Statement on the "AI pause" letter, DAIR Institute, 2023](https://www.dair-institute.org/blog/letter-statement-March2023/) How this view has changed: The view predates the current debate. In a 2016 Facebook post after NIPS she wrote that she was not worried about machines taking over the world but about "groupthink, insularity, and arrogance in the AI community." By September 2026 she was calling the extinction story a "machine-god narrative" that is "meant to distract us." ### Artificial general intelligence Editorial summary: The pursuit of AGI, in her analysis, descends from eugenics-era ideas about improving humanity, and an unspecified system cannot be tested for safety; the alternative is to build scoped systems with testable protocols. > undefined systems like 'AGI' cannot be appropriately tested for safety Source: [The TESCREAL bundle, First Monday, 2024](https://firstmonday.org/ojs/index.php/fm/article/view/13636) ### AI hype and anthropomorphism Editorial summary: Chatbots are built to make users believe a mind sits behind the text; she wants researchers and the public to "unlearn" that framing and to judge systems as engineered products with owners. > These chatbots, like Claude or ChatGPT, are designed to make you believe that there's some sort of superhuman brain behind whatever is outputting these texts. Source: [Interview on Democracy Now!, 2026](https://www.democracynow.org/2026/8/13/timnit_gebru) How this view has changed: In DAIR's 2021 launch statement the complaint was that AI had been raised to a superhuman level that made it seem "both inevitable and beyond our control"; by 2026 she describes the anthropomorphism as a deliberate product design choice. ### Data and labor Editorial summary: Large models rest on web-scale data taken without consent and on data labelers paid close to nothing; DAIR's Data Workers' Inquiry exists to let those workers document and organize around their own conditions. > These people have to painstakingly supply the data and label it. And they work in overwhelmingly exploitative conditions. They are paid maybe $1 an hour. Source: [Interview on Democracy Now!, 2026](https://www.democracynow.org/2026/8/13/timnit_gebru) ### Scale and large language models Editorial summary: Bigger models carry environmental and financial costs, encode the dominant views of whoever is on the internet, and generate text fluent enough to be mistaken for understanding. > The internet represents hegemonic views. It does not represent views of everybody in the world. Source: [Interview on Democracy Now!, 2026](https://www.democracynow.org/2026/8/13/timnit_gebru) ### Independence of research from Big Tech Editorial summary: Ethics research cannot flourish inside institutions that punish it; the fix is structural, starting with who pays for the science. > I think that, number one, the incentive needs to be decoupled from corporations or the military. Source: [Scientific American, 2026](https://www.scientificamerican.com/article/timnit-gebru/) How this view has changed: In 2022 she told Stanford HAI that "this needs to be about institutional and structural change"; by 2026 the prescription had narrowed to funding, which she called "the number-one thing." ### Bias is not only a data problem Editorial summary: When Yann LeCun said in June 2020 that machine learning systems are biased because their data is biased, Gebru objected that harms cannot be reduced to dataset composition; they also come from problem framing, deployment and who is in the room. > I'm sick of this framing. Tired of it. Many people have tried to explain, many scholars. Listen to us. You can't just reduce harms caused by ML to dataset bias. Source: [Synced, reporting the exchange on Twitter, 2020](https://syncedreview.com/2020/06/30/yann-lecun-quits-twitter-amid-acrimonious-exchanges-on-ai-bias/) ### Facial recognition and surveillance Editorial summary: Facial analysis systems fail most on the people most exposed to surveillance, and she has argued since 2019 that police should not buy them. > darker-skinned females are the most misclassified group (with error rates of up to 34.7%). The maximum error rate for lighter-skinned males is 0.8%. Source: [Gender Shades, FAT\* 2018, 2018](https://proceedings.mlr.press/v81/buolamwini18a.html) ## Critics and counterpoints ### Existential risk versus present harms Their view: Extinction scenarios come from a longtermist ideology and are, in her words, "meant to distract us" from documented harms such as exploited data workers, surveillance, autonomous weapons and the climate cost of training. The case against: Geoffrey Hinton, asked on CNN in May 2023 why he had not defended Gebru at Google, said her concerns "weren't as existentially serious" as machines surpassing and displacing humans. Yoshua Bengio has written that he is asked to stop discussing catastrophic risk because it would "suck all the air out of the room," and answers that democracies routinely debate long-term and short-term dangers at once, that plausible arguments exist for a superhuman system acquiring self-preservation goals, and that even a low-probability catastrophe of that size "should rationally demand our attention." Where it stands: Hinton, Bengio, Altman, Amodei and Hassabis signed the May 2023 Center for AI Safety statement; Gebru did not, and in September 2026 was still calling the extinction story a "machine-god narrative." Both camps now say the other is serving industry interests. [Source](https://yoshuabengio.org/en/blog/reasoning-through-arguments-against-taking-ai-safety-seriously) ### Is algorithmic bias a data problem? Their view: Dataset composition is one cause among many; harms also come from how a problem is framed, who builds the system, how it is deployed and on whom, so "fix the data" is not a sufficient response. The case against: In June 2020 Yann LeCun wrote that "ML systems are biased when data is biased," using a face-upsampling model that turned Barack Obama's pixelated photo white because it was pretrained on a mostly white face dataset, and that training on different data would give different results. The steelman is that this is literally true of that model and that engineers need actionable causes, not sociology. Where it stands: Gebru replied that she was "sick of this framing"; after several days of argument LeCun left Twitter on 28 June 2020, asking people to stop attacking Gebru and his critics alike. The exchange is still cited as the moment the fairness field's technical and structural wings split in public. [Source](https://syncedreview.com/2020/06/30/yann-lecun-quits-twitter-amid-acrimonious-exchanges-on-ai-bias/) ### Can language models be too big? Their view: Scale carries environmental, financial and social costs that fall on people who never benefit, and it crowds out smaller, better-understood approaches; documentation and scoped systems should come before size. The case against: Dario Amodei's October 2024 essay argues that powerful AI could arrive as early as 2026 and compress decades of biological and medical progress into a few years, gains he holds justify the costs of scaling. Sam Altman wrote in 2025 that "we are past the event horizon; the takeoff has started" and that the productivity gains will be "enormous." Both treat the costs Gebru lists as real but secondary. Where it stands: The labs kept scaling, and by 2026 Gebru was arguing that the same companies' doom warnings and capability claims are one marketing story; her preferred alternative, task-specific models built by small groups, is what DAIR's Huniki Federation funds. [Source](https://darioamodei.com/machines-of-loving-grace) ### Was she fired or did she resign? Their view: She set conditions for staying, offered to work out a last date if they were not met, and was instead cut off while on leave; that is a firing, and the paper was the pretext. The case against: Jeff Dean's memo to Google Research says the paper was shared "with a day's notice" against a two-week review requirement, that reviewers found it "didn't meet our bar for publication," and that when she made continued employment conditional on demands the company would not meet, "we accept and respect her decision to resign." Google has never changed that position. Where it stands: Sundar Pichai apologized to staff for how the departure had "seeded doubts" and ordered a review, but did not say Google erred in removing her. Wired's June 2021 reconstruction, drawing on both sides, reports that Kacholia's email told Gebru her employment would end "faster than your email reflects." [Source](https://www.platformer.news/the-withering-email-that-got-an-ethical/) ### Should ethics research live inside the companies it studies? Their view: It cannot, or not reliably: her own team was disbanded over one paper, and the incentives of corporate research reward speed and product alignment over scrutiny; funding must come from elsewhere. The case against: The case for inside work, made by her former Google colleagues and by Google itself after 2021, is that internal teams see systems before release, can change products directly, and that Google responded to the episode by tying executive pay to diversity progress and giving researchers clearer guidance on sensitive topics. Her adviser Fei-Fei Li's Stanford HAI takes a middle path, accepting industry funding while arguing for academic independence. Where it stands: Margaret Mitchell was fired two months after Gebru; Alex Hanna and Dylan Baker left Google for DAIR; Gebru told Scientific American in June 2026 that the answer is to decouple incentives from corporations and the military. [Source](https://www.scientificamerican.com/article/timnit-gebru/) ## Misconceptions Claim: Google fired her because Stochastic Parrots failed peer review. Correction: The paper was accepted and published at FAccT in March 2021. The objection came from Google's internal review, which the company says found it "didn't meet our bar for publication"; the exit followed her conditions for staying, which Google treated as a resignation and she treats as a firing. [Source](https://www.platformer.news/the-withering-email-that-got-an-ethical/) Claim: Gebru is against AI. Correction: She is an electrical engineer who built Apple audio hardware and trained the car detectors in her own PhD; DAIR builds task-specific language models for East African languages and funds community groups doing the same. Her target is scale-at-any-cost and the claims made for it, not the technology. [Source](https://www.dair-institute.org/projects/) Claim: Gender Shades tested facial recognition. Correction: It tested gender classification, a different task: three commercial systems were asked whether a face was male or female. The result is often cited in debates about identification systems, which is why the 2019 Amazon letter had to argue separately about Rekognition. [Source](https://proceedings.mlr.press/v81/buolamwini18a.html) Claim: Her memoir is called The View from Somewhere. Correction: That was the working title NISO used in February 2025. The book was retitled Deep Unlearning: The Rise of AI and the Radicalization of a Tech Idealist and is scheduled for 16 February 2027. [Source](https://brittlepaper.com/2026/08/ethiopian-born-computer-scientist-timnit-gebrus-debut-book-deep-unlearning-is-now-available-for-preorder/) ## Quotes > One of the most important emergent issues plaguing our society today is that of algorithmic bias. Most works based on data mining, including my own works described in this thesis, suffer from this problem Source: [Final chapter of her Stanford thesis, quoted in Wired, 2017](https://www.wired.com/story/google-timnit-gebru-ai-what-really-happened/) > Have you ever heard of someone getting 'feedback' on a paper through a privileged and confidential document to HR? Source: [Email to the Google Brain Women and Allies list, published by Platformer, 2020](https://www.platformer.news/the-withering-email-that-got-an-ethical/) > They would never do what they did to me to someone else. Google had a problem with me speaking up about discrimination. Source: [Nature's 10, 2021](https://www.nature.com/articles/d41586-021-03621-0) > AI needs to be brought back down to earth. It has been elevated to a superhuman level that leads us to believe it is both inevitable and beyond our control. Source: [DAIR launch press release, 2021](https://dair-institute.org/press-release/) > We want to do interdisciplinary research. We don't want to drive people to the publishing rat race. Source: [IEEE Spectrum, 2022](https://spectrum.ieee.org/timnit-gebru-dair-ai-ethics) > What I've realized is that we can talk about the ethics and fairness of AI all we want, but if our institutions don't allow for this kind of work to take place, then it won't. Source: [Stanford HAI, 2022](https://hai.stanford.edu/news/timnit-gebru-ethical-ai-requires-institutional-and-structural-change) > Question everything, question the source of information that you get. Don't believe any kind of hype that you hear from companies. Source: [Scientific American, 2026](https://www.scientificamerican.com/article/timnit-gebru/) > When a bridge collapses, you don't analyze whether the bridge was ethical or sentient or why it decided to collapse. You ask, who is the person who built this bridge to be so flimsy? Source: [Interview with Wired, 2026](https://www.wired.com/story/one-of-ais-fiercest-critics-says-all-the-doom-talk-is-meant-to-distract-us/) ## Related leaders - [fei-fei-li](https://frontierminds.ai/people/fei-fei-li/): Gebru's PhD adviser at Stanford and senior author on the 2017 PNAS Street View paper; Gebru has since argued that AI research must be decoupled from the corporate funding that Li's Stanford HAI accepts. - [geoffrey-hinton](https://frontierminds.ai/people/geoffrey-hinton/): Asked on CNN in May 2023 why he had not spoken up for Gebru at Google, Hinton said her concerns "weren't as existentially serious" as machines surpassing humans; the two are the poles of the present-harms versus extinction-risk debate. - [yoshua-bengio](https://frontierminds.ai/people/yoshua-bengio/): Co-signatory with Gebru of the April 2019 letter asking Amazon to stop selling facial recognition to police; four years later he signed the pause letter that Gebru and her co-authors rebutted, and his 2024 essay answers her line of argument directly. - [yann-lecun](https://frontierminds.ai/people/yann-lecun/): Their June 2020 exchange over whether bias is only a data problem ended with LeCun leaving Twitter; they remain opposed on that question and agree, from different directions, that extinction risk is overstated. - [dario-amodei](https://frontierminds.ai/people/dario-amodei/): Critic of the scaling case Amodei makes in "Machines of Loving Grace"; when he called in September 2026 for the industry to "pace the frontier," she told CNBC that "There is no AI exemption in existing statutes." - [sam-altman](https://frontierminds.ai/people/sam-altman/): Her TESCREAL framework with Emile P. Torres underpins the critique of OpenAI's AGI mission; she cites OpenAI's 2019 claim that GPT-2 was too dangerous to release as evidence that hype and doom are the same marketing. - [emily-bender](https://frontierminds.ai/people/emily-bender/): Co-author of "On the Dangers of Stochastic Parrots" (2021), whose title phrase Bender coined, and of the March 2023 statement on the pause letter; her podcast with Alex Hanna, Mystery AI Hype Theater 3000, is produced with DAIR. - [mustafa-suleyman](https://frontierminds.ai/people/mustafa-suleyman/): Business Insider's 2021 report on his move to a Google vice-presidency, after complaints about his management at DeepMind, contrasted it with Google's ousting of Gebru from its Ethical AI team months earlier. ## Affiliations - Distributed AI Research Institute (Founder and Executive Director, 2021-) - Black in AI (Co-founder, 2017) - AddisCoder (Board member) - Google (Research Scientist and co-lead, Ethical AI team, 2018-2020) - Microsoft Research, FATE group, New York (Postdoctoral Researcher, 2017-2018) - Stanford Artificial Intelligence Laboratory (PhD, 2017) - Apple (Audio hardware and signal processing engineer) ## Areas of focus - algorithmic bias - data documentation and governance - AI accountability and auditing - data labor - critique of AGI and longtermism - community-rooted research ## Tags - AI ethics - bias - data governance - accountability - AI hype ## Organizations - dair-institute: Founder and Executive Director (2021–present) - stanford-university: PhD, Stanford Artificial Intelligence Laboratory (2013–2017) ## Links - [DAIR Institute](https://www.dair-institute.org/) - [DAIR profile](https://dair-institute.org/team/timnit-gebru/) - [Stanford homepage](https://ai.stanford.edu/~tgebru/) - [Google Scholar](https://scholar.google.com/citations?user=lemnAcwAAAAJ) - [x](https://x.com/timnitGebru) - [mastodon](https://dair-community.social/@timnitGebru) - [linkedin](https://www.linkedin.com/in/timnit-gebru-7b3b407/) - [website](https://ai.stanford.edu/~tgebru/) ## Sources 1. [What Really Happened When Google Ousted Timnit Gebru - Wired (8 June 2021)](https://www.wired.com/story/google-timnit-gebru-ai-what-really-happened/) 2. [The withering email that got an ethical AI researcher fired at Google, with Jeff Dean's response - Platformer (3 December 2020)](https://www.platformer.news/the-withering-email-that-got-an-ethical/) 3. [Google CEO pledges to investigate exit of top AI ethicist, with Sundar Pichai's memo - Axios (9 December 2020, Wayback Machine copy)](https://web.archive.org/web/20260106235932/https://www.axios.com/2020/12/09/sundar-pichai-memo-timnit-gebru-exit) 4. [We read the paper that forced Timnit Gebru out of Google - MIT Technology Review (4 December 2020)](https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru/) 5. [Timnit Gebru: AI ethics leader - Nature's 10 (December 2021)](https://www.nature.com/articles/d41586-021-03621-0) 6. [Press Release - Announcing DAIR (2 December 2021)](https://dair-institute.org/press-release/) 7. [After being pushed out of Google, Timnit Gebru forms her own AI research institute - TechCrunch (2 December 2021)](https://techcrunch.com/2021/12/02/google-timnit-gebru-ai-research-dair/) 8. [The Distributed AI Research Institute joins CS&S - Code for Science and Society (2 December 2021)](https://www.codeforsociety.org/fsp/news/the-distributed-ai-research-institute-joins-cs-s) 9. [Timnit Gebru Is Building a Slow AI Movement - IEEE Spectrum (31 March 2022)](https://spectrum.ieee.org/timnit-gebru-dair-ai-ethics) 10. [Timnit Gebru: Ethical AI Requires Institutional and Structural Change - Stanford HAI (26 May 2022)](https://hai.stanford.edu/news/timnit-gebru-ethical-ai-requires-institutional-and-structural-change) 11. [Statement from the listed authors of Stochastic Parrots on the "AI pause" letter - DAIR (31 March 2023)](https://www.dair-institute.org/blog/letter-statement-March2023/) 12. [Pause Giant AI Experiments - An Open Letter - Future of Life Institute (22 March 2023)](https://futureoflife.org/open-letter/pause-giant-ai-experiments/) 13. [Statement on AI Risk - Center for AI Safety (May 2023)](https://safe.ai/work/statement-on-ai-extinction-risk) 14. [The TESCREAL bundle - First Monday 29(4) (April 2024)](https://firstmonday.org/ojs/index.php/fm/article/view/13636) 15. [Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification - PMLR (2018)](https://proceedings.mlr.press/v81/buolamwini18a.html) 16. [Datasheets for Datasets - arXiv (v1 March 2018, v8 December 2021)](https://arxiv.org/abs/1803.09010) 17. [Model Cards for Model Reporting - arXiv (2018/2019)](https://arxiv.org/abs/1810.03993) 18. [Closing the AI Accountability Gap - arXiv (January 2020)](https://arxiv.org/abs/2001.00973) 19. [Using Deep Learning and Google Street View to Estimate the Demographic Makeup of the US - arXiv (February 2017)](https://arxiv.org/abs/1702.06683) 20. [On the Dangers of Stochastic Parrots - ACM Digital Library (FAccT 2021)](https://dl.acm.org/doi/10.1145/3442188.3445922) 21. [Model Cards - Hugging Face Hub documentation](https://huggingface.co/docs/hub/model-cards) 22. [Timnit Gebru - Stanford AI Lab homepage](https://ai.stanford.edu/~tgebru/) 23. [Timnit Gebru - Google Scholar profile (checked September 2026)](https://scholar.google.com/citations?user=lemnAcwAAAAJ&hl=en) 24. [About - Black in AI](https://www.blackinai.org/about) 25. [Meet DAIR - Timnit Gebru; Team; Projects; Blog - DAIR Institute](https://dair-institute.org/team/timnit-gebru/) 26. [Projects - DAIR Institute](https://dair-institute.org/projects/) 27. [Data Workers' Inquiry](https://data-workers.org/) 28. [AI researchers tell Amazon to stop selling 'flawed' facial recognition to the police - The Verge (3 April 2019)](https://www.theverge.com/2019/4/3/18291995/amazon-facial-recognition-technology-rekognition-police-ai-researchers-ban-flawed) 29. [Yann LeCun Quits Twitter Amid Acrimonious Exchanges on AI Bias - Synced (30 June 2020)](https://syncedreview.com/2020/06/30/yann-lecun-quits-twitter-amid-acrimonious-exchanges-on-ai-bias/) 30. [Geoffrey Hinton's misguided views on AI - Paris Marx, Disconnect (quoting Hinton's May 2023 CNN interview)](https://disconnect.blog/geoffrey-hintons-misguided-views-on-ai/) 31. [Reasoning through arguments against taking AI safety seriously - Yoshua Bengio (9 July 2024)](https://yoshuabengio.org/en/blog/reasoning-through-arguments-against-taking-ai-safety-seriously) 32. [Machines of Loving Grace - Dario Amodei (October 2024)](https://darioamodei.com/machines-of-loving-grace) 33. [The Gentle Singularity - Sam Altman (2025)](https://blog.samaltman.com/the-gentle-singularity) 34. [Timnit Gebru - The 100 Most Influential People of 2022 - TIME](https://time.com/collections/100-most-influential-people-2022/6177822/timnit-gebru/) 35. [Timnit Gebru - Fortune's World's 50 Greatest Leaders 2021](https://fortune.com/ranking/worlds-greatest-leaders/2021/timnit-gebru/) 36. [Timnit Gebru - Great Immigrants 2023 - Carnegie Corporation of New York](https://www.carnegie.org/awards/honoree/timnit-gebru/) 37. [BBC 100 Women 2023 - BBC](https://www.bbc.co.uk/news/resources/idt-02d9060e-15dc-426c-bfe0-86a6437e5234) 38. [Dr. Timnit Gebru Is Our 2025 Miles Conrad Awardee - NISO (February 2025)](https://www.niso.org/press-releases/dr-timnit-gebru-our-2025-miles-conrad-awardee) 39. [Timnit Gebru on how to safeguard independent science for the AI age - Scientific American (16 June 2026)](https://www.scientificamerican.com/article/timnit-gebru/) 40. ["Deep Unlearning": Timnit Gebru on AI Hype, Ethics and Algorithmic Racial Bias - Democracy Now! (13 August 2026)](https://www.democracynow.org/2026/8/13/timnit_gebru) 41. [Timnit Gebru's Debut Book, Deep Unlearning, Is Now Available for Preorder - Brittle Paper (26 August 2026)](https://brittlepaper.com/2026/08/ethiopian-born-computer-scientist-timnit-gebrus-debut-book-deep-unlearning-is-now-available-for-preorder/) 42. [Deep Unlearning - Apple Books listing (Atria/One Signal, 16 February 2027)](https://books.apple.com/us/book/deep-unlearning/id6780517958) 43. [One of AI's Fiercest Critics Says All the Doom Talk Is 'Meant to Distract Us' - Wired (11 September 2026)](https://www.wired.com/story/one-of-ais-fiercest-critics-says-all-the-doom-talk-is-meant-to-distract-us/) 44. [Trump opposition to AI rules undercuts industry's calls for a slowdown - CNBC (15 September 2026, with Gebru's response to Amodei)](https://www.cnbc.com/2026/09/15/trump-opposition-to-ai-rules-undercuts-industrys-calls-for-a-slowdown.html) 45. [The First Keynote and Featured Sessions for SXSW 2027 Are Here - SXSW (15 September 2026)](https://sxsw.com/2026/the-first-keynote-and-featured-sessions-for-sxsw-2027-are-here/) 46. [Timnit Gebru - Wikipedia](https://en.wikipedia.org/wiki/Timnit_Gebru) 47. [File - Timnit Gebru crop.jpg - Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Timnit_Gebru_crop.jpg) ## Portrait credit AI-generated watercolor interpretation based on a reference photograph. [Photo: TechCrunch, CC BY 2.0, via Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Timnit_Gebru_crop.jpg) [CC BY 2.0](https://creativecommons.org/licenses/by/2.0/)