Timnit Gebru is an Ethiopian-American computer scientist and founder of the Distributed AI Research (DAIR) Institute. She previously co-led Google's Ethical AI team until her departure in December 2020, which she has described as a firing and which followed a dispute over a paper examining the risks of large language models. The episode became a reference point in debates over corporate control of AI ethics research and the limits of conducting accountability work inside the companies being held to account.
Research
Two of Gebru's papers are widely cited in the AI fairness and language-model-risk literatures.
Gender Shades (2018), co-authored with Joy Buolamwini, audited commercial facial-analysis systems and reported higher error rates for darker-skinned women than for lighter-skinned men. The study is frequently cited as a demonstration that deployed AI systems encode demographic bias, and it informed subsequent regulation and corporate policy on facial recognition.
"On the Dangers of Stochastic Parrots" (2021), co-authored with Emily Bender, Margaret Mitchell, and Angelina McMillan-Major, argued that ever-larger language models carry under-examined environmental, financial, and social costs, and that scale does not guarantee understanding or benefit. It was the paper at the center of her exit from Google and is a frequently cited reference in the literature on language-model risk.
DAIR Institute
In 2021 Gebru founded DAIR, described as a "globally distributed organization of academics, activists, and engineers conducting community-rooted research" independent of Big Tech funding and direction (Source: https://dair-institute.org/). DAIR was launched with support from the MacArthur, Ford, and Kapor foundations and Open Society (Source: https://www.macfound.org/press/grantee-news/new-institute-to-focus-on-community-rooted-ai-research). Its stated premise is that AI research framed around, and accountable to, the communities most affected by AI systems produces different questions than research conducted inside the firms building the technology.
Positions on AI policy
Gebru advocates centering AI governance on present-day harms — bias, discrimination, labor exploitation, surveillance, and environmental cost — rather than on speculative existential risk. She is a central figure in the AI Safety vs. AI Ethics Divide debate, arguing that the safety community's focus on future superintelligence diverts attention and resources from communities already harmed by deployed systems. She and Émile Torres have labelled a cluster of ideologies "TESCREAL", arguing that an AGI-centric framing carries its own ideological commitments. Her work is associated with the fairness and accountability wing of AI Ethics Frameworks and has shaped Algorithmic Accountability and Bias Audits practice.
Following her departure from Google, Gebru founded the Distributed AI Research Institute (2021), on the premise that research into the harms of AI systems cannot be conducted reliably inside the companies building them.
Relationships
- anchors: AI Ethics Frameworks — a founding figure of the fairness/accountability tradition
- collaborates-with: Joy Buolamwini — co-author of Gender Shades
- related: AI Now Institute — aligned on present-harms framing of AI governance
- related: AI Safety vs. AI Ethics Divide — central figure in this debate
- supports: Algorithmic Accountability and Bias Audits — advocate for accountability-first governance