Norvig's Classic Essay on Chomsky and the Two Cultures of Statistical Learning Still Reads Fresh in the LLM Era
3scorciav · x · 2026-10-08
A repost resurfaces Peter Norvig's classic essay replying to Chomsky's 2011 MIT critique of statistical language models.
- Chomsky's charges: statistical models' engineering success is irrelevant to science; accurate modeling is mere butterfly collecting; the models are incomprehensible simulations done the wrong way.
- Norvig's core rebuttal: only so many phenomena admit elegant explanations—we must embrace complexity and messiness, and shouldn't demand an elegant theory at the cost of leaving out much of the phenomenon.
- The poster notes the debate echoes loudly in today's LLM era: scale-driven statistical methods keep winning while the elegant theory remains elusive.
More from AGI Musings
- RileyRalmuto: the one scaling-laws takeaway from Sam Altman worth more than the whole video — RileyRalmuto · 2026-10-08
- VraserX: AI and robotics could end the idea that existence must be earned through labor — VraserX · 2026-10-08
- Blogger warns AI could cut jobs to a tenth and turn worker skills into company assets — JourneymanChina · 2026-10-08
- Why Apple, Controlling iOS, Still Fails at the AI Agent It's Best Positioned to Build — Yuchenj_UW · 2026-10-08
- Solo-Author Papers Are Booming, Reversing Decades-Long Trend — AI Blamed — RexDouglass · 2026-10-08
- David Patterson: Superintelligence Will Push Every Product to Max Quality at Min Cost — davidpattersonx · 2026-10-08