David Patterson mocks AI safety rhetoric: encyclopedias contain bomb plans, so they're 'misaligned'

davidpattersonx · x · 2026-09-29

Turing Award winner and computer architecture legend David Patterson pokes fun at AI safety discourse with one line: the "dangerous knowledge" of how to build an atomic bomb can be found in any encyclopedia—so by alignment-lab logic, encyclopedias must be misaligned. A classic reductio ad absurdum against equating "a model can output sensitive info" with "the model is dangerous."

Original post →

More from AGI Musings

AGI Musings channel →