Meta's Scientific AI Was Shut Down After Just 3 Days for Generating Dangerous Misinformation
In November 2022, Meta launched Galactica — an AI model trained on 48 million scientific papers, touted as a tool that would 'summarize academic literature, solve math problems, and annotate proteins.' Within 72 hours, researchers were posting examples of it confidently generating fake citations, inventing nonexistent studies, producing racist content about the history of bears, and writing authoritative-sounding nonsense about the benefits of eating glass. Meta pulled the public demo 3 days after launch.
In November 2022, Meta launched Galactica — an AI model trained on 48 million scientific papers, touted as a tool that would 'summarize academic literature, solve math problems, and annotate proteins.' Within 72 hours, researchers were posting examples of it confidently generating fake citations, inventing nonexistent studies, producing racist content about the history of bears, and writing authoritative-sounding nonsense about the benefits of eating glass. Meta pulled the public demo 3 days after launch.
Weirdness Classification
9/10 — Deeply unhinged
Field Reports (0)
Loading reports...
Sign in to file your field report.
Know something weirder?
Submit your own AI incident report to the public record.