"Stochastic Parrot" Feud Resurfaces as LLMs Outdo Humans at Frontier Math; Oversight Debate Rekindled
ctjlewis · x · 2026-09-23
- @basedjensen mocks the classic "stochastic parrot" critique, arguing the parrot now does frontier math better than humans and that the paper's authors are the real parrots.
- @mmitchellai pushes back with a design point: because agents don't reliably do what we intend via natural language, building oversight mechanisms arguably matters more right now than expanding agent action surfaces — citing oversight work from Eric Horvitz, Saleema Amershi and others, with a linked paper.
More from AGI Musings
- Google researcher's refugee childhood story pushback on the teen drop-out-to-SF founder narrative — satnam6502 · 2026-09-23
- Miles Brundage worries AI info flood is breaking our cognitive defenses — Miles_Brundage · 2026-09-23
- AI safety researcher David Krueger: CEO doom claims lack credibility, but transparency research is badly needed — DavidSKrueger · 2026-09-23
- AI Safety Researcher David Krueger on Extinction Risk, CEOs and Regulation — DavidSKrueger · 2026-09-23
- AI slop reveals what hiring and performance reviews should actually measure — LadyAshBorg · 2026-09-23
- AI use in math papers jumped from 1.39% to 14.09% in six months, arXiv study finds — rohanpaul_ai · 2026-09-23