An independent AI research institute founded in 2021 by Timnit Gebru, following her departure from Google amid the dispute over "On the Dangers of Stochastic Parrots". Its organising premise is that research on the harms of AI systems cannot be conducted reliably inside the companies building them, and that it should be distributed rather than concentrated in a single institution or region.
Research posture
DAIR's programme continues the line of argument the Stochastic Parrots paper set out: that the salient questions about large language models concern documented, present harms — the distribution of environmental and financial costs, what uncurated web-scale training data encodes, opportunity cost in the research field, and the accountability gap created by fluent synthetic text — rather than speculative future capability.
The paper's closing position remains the programme's clearest statement: "applications that aim to believably mimic humans bring risk of extreme harms. Work on synthetic human behavior is a bright line in ethical AI development." That formulation ties DAIR's work to the stochastic parrots critique and distinguishes it from the existential-risk research programme, which locates the concern in future systems rather than in deployed ones.
Relationships
- related: Timnit Gebru — founder
- related: On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? (Bender, Gebru, McMillan-Major, Shmitchell — FAccT 2021), Stochastic Parrots (Bender, Gebru, McMillan-Major, Mitchell, 2021) and the Octopus Test, Emily M. Bender, AI Bias and Discrimination