# DAIR Institute (Distributed AI Research) **Draft profile.** Independent AI research institute founded by **[[Timnit Gebru]]** in **December 2021** as *"an independent, community-rooted institute set to counter Big Tech's pervasive influence on the research, development and deployment of AI."* Headquartered in **Oakland, California**, with researchers distributed across Africa, North America, and Europe. Launched with **$3.7M** in initial funding from a coalition of progressive and traditional foundations: **[[Ford Foundation]], [[MacArthur Foundation]], [[Kapor Center]], Open Society Foundation, and [[Rockefeller Foundation]]**. **Critically important for this KB:** DAIR is the **structural counter-model to the [[TESCREAL]] AI-safety institutional stack** — the [[MIRI]] / [[Future of Life Institute]] / [[Anthropic]] / [[Open Philanthropy]] AI-safety operation. Where the TESCREAL stack frames AI risk as **future / existential / civilizational** and funds technical alignment research, DAIR frames AI risk as **present / harm-based / community-affected** and funds research grounded in the communities most impacted by AI systems. ## Origin Timnit Gebru was **fired from Google in December 2020** for raising concerns about discrimination in AI systems and refusing to suppress her research on the dangers of large language models (the "[[Stochastic Parrots]]" paper she co-authored became one of the most-cited critiques of LLM development). The firing triggered international academic outrage, internal Google revolt, and a wave of high-profile resignations. DAIR was launched exactly **one year later**, on December 2, 2021. The timing was deliberate — Gebru framed DAIR as the institutional response to the experience: an AI ethics operation that **couldn't be fired** because it answered to a community rather than to a corporate parent. ## What DAIR Does ### Research focus DAIR's research centers on **documenting real harms** of AI systems rather than speculating about hypothetical future risks: - **Algorithmic bias** in deployed AI systems (hiring, criminal justice, healthcare). - **Labor exploitation** in AI supply chains (data labeling, content moderation). - **Environmental impact** of large-model training. - **Surveillance applications** of AI in marginalized communities. - **AI hype critique** — debunking inflated claims about AI capabilities. ### Methodological commitments - **Community-rooted** — research conducted with and by affected communities, not just about them. - **Globally distributed** — researchers across Africa, North America, Europe. - **Anti-burnout** — explicit institutional commitment to reasonable workloads, opposing what Gebru calls *"discriminatory exploitation"* in academic AI work. - **Slow publishing** — guidelines that encourage prioritizing health over publishing volume, in deliberate contrast to the AI-research field's burnout culture. ## Why It Matters in This KB ### 1. It is the institutional embodiment of the TESCREAL critique [[Timnit Gebru]] and [[Émile Torres]] co-coined the term **[[TESCREAL]]** to describe the ideological bundle (Transhumanism, Extropianism, Singularitarianism, Cosmism, Rationalism, [[Effective Altruism]], Longtermism). DAIR is what happens when one of the two coiners builds a research institution explicitly **to operate outside the TESCREAL framework**. **The existence of DAIR is itself an argument** — proof that AI ethics research can be funded, conducted, and published from outside the EA/longtermist orbit. ### 2. It demonstrates the present-harm vs. future-risk fault line The TESCREAL stack ([[MIRI]], [[FLI]], [[Anthropic]] safety teams, much of [[Open Philanthropy]]'s AI grantmaking) frames AI risk in **future / existential / civilizational** terms — alignment of hypothetical superintelligences, catastrophic misuse scenarios, civilizational-trajectory questions. DAIR frames AI risk in **present / documented / community-affected** terms — biased models *now* harming *specific* people. **These are different research programs producing different policy implications**, and the philanthropic capital underwriting each side flows from largely non-overlapping donor networks. ### 3. The funder composition is significant DAIR's launch funders — **Ford, MacArthur, Kapor, Open Society, Rockefeller** — are the **traditional progressive-foundation coalition**, almost entirely absent from the [[Open Philanthropy]] / [[Schmidt Futures]] / [[Survival and Flourishing Fund]] AI-safety funding stack. This is one of the cleanest documented cases of **two parallel philanthropic operations funding diametrically opposed approaches to the same problem domain** with no donor overlap. ### 4. It exists as a critique of philanthropy's role in AI ethics Gebru has publicly criticized large philanthropic involvement in AI research as a mechanism by which billionaires shape "ethics" discourse in self-serving ways. DAIR taking funding from traditional progressive foundations rather than from EA megafunders is itself a structural commentary on which donor networks can credibly underwrite ethics work. ## Critiques and Defenses ### From the TESCREAL / EA side - DAIR's "present-harm" focus is sometimes characterized as **neglecting catastrophic / existential** risks that could affect vastly more future people. - Some EA-aligned critics argue DAIR's progressive-foundation funding base creates its own ideological selection effects. ### From DAIR / Gebru's perspective - "Existential risk" framing serves to **distract from documented present harms** and concentrates AI policy in the hands of the same companies causing those harms. - The longtermist framing is **structurally rigged** to favor concentration of decision-making in technocratic elites — exactly the [[TESCREAL]] critique Gebru and [[Émile Torres]] formalized. ### From traditional AI research - DAIR's small size (~10–15 researchers) limits research throughput compared to the much-larger TESCREAL-coded operations. ## Draft Notes / To Expand - Document specific DAIR research outputs (data-labor papers, large-language-model critiques, surveillance documentation) - Map [[Stochastic Parrots]] paper and its institutional aftermath - Add the connection to [[Black in AI]] and adjacent equity-in-AI organizations - Detail the relationship between [[Timnit Gebru]]'s Google firing and the DAIR founding - Trace the [[Émile Torres]] / Gebru collaboration on TESCREAL framework - Investigate DAIR's policy engagement (FTC, Congress, EU AI Act consultations) - Document specific donor relationships and grant terms ## See Also - [[Timnit Gebru]] — Founder and director - [[Émile Torres]] — TESCREAL framework co-creator - [[TESCREAL]] — The ideological bundle DAIR operates outside of - [[Stochastic Parrots]] (currently unlinked) — Foundational paper - [[MIRI]] — The closest TESCREAL-coded structural counterpart - [[Open Philanthropy]] — TESCREAL-aligned mega-funder DAIR critiques - [[Anthropic]] — Major AI lab whose safety work DAIR critiques - [[Philanthropy Ecosystem]] — Money-flow MOC - [[AI Safety]] — Adjacent concept (DAIR's framing differs from the standard usage)