Any finite state machine could be a 'stochastic parrot', researcher rebuts the analogy

aran_nayebi · x · 2026-09-27

Researcher Aran Nayebi responds to Timnit Gebru's "stochastic parrot" framing, arguing the analogy doesn't track capability endowed by training. His point: any finite state machine — including the human brain — could be labeled a "stochastic parrot" purely because of how probabilities factorize, which tells you nothing about its actual intelligence. The analogy thus carries no informational weight for judging model capability.

Related event: Researchers Push Back on 'Stochastic Parrot' Claim as LLMs Crack Hard Math(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →