Timnit Gebru, former co‑lead of Google’s Ethical AI team, was dismissed in December 2020 after refusing to retract the paper *On the Dangers of Stochastic Parrots*. The brief 14‑page study warned that massive language models would become fluent parrots, hallucinating facts while lacking true understanding.

Since the paper’s release, every prediction has surfaced. Amazon’s hiring algorithm flagged resumes containing “women,” hospitals’ risk scores underestimated Black patients, and Apple Card’s credit limit for wives fell to one‑tenth of husbands’ for identical profiles. These cases confirm the bias‑amplification claim.

Environmental estimates in the study equated a single model’s training to five cars’ lifetime emissions. Google’s 2024 emissions rose 48 % from 2019, largely due to AI workloads, and Microsoft followed with a 29 % increase. This aligns with recent reports that large‑scale training consumes as much energy as small nations. The paper’s carbon warning now matches corporate data.

Auditability gaps also proved accurate. The LAION‑5B dataset, used for Stable Diffusion, contained child‑abuse material that companies could not detect. Gebru’s call for transparent data supply chains has emerged as a central issue for responsible model training. The industry now faces the very risks she highlighted.