Hinton: new-language experiment shows LLMs truly understand, not just predict next word
coherence · x · 2026-11-20
Geoffrey Hinton pushes back on the claim that LLMs are "just predicting the next word," arguing it falls apart when models must understand something they've never seen.
He cites an experiment by a former graduate student who defined a new programming language with unusual syntax, gave a model a program in it, and asked what the program would do — the model answered correctly. Hinton says correctly interpreting a novel language definition requires genuine understanding, not mere next-token prediction.
Related event: Hinton Pushes Back on "LLMs Just Predict the Next Word"(2 posts)→
More from AGI Musings
- 25 exec calls fed to Claude reveal the 10 real reasons enterprise AI transformation stalls — vasuman · 2026-11-20
- Reid Hoffman: AI's real superpower for founders is speed, not just IQ — reidhoffman · 2026-11-20
- Biocomputing turns living human neurons into AI — and raises the question of suffering — frugal-tech-worker · 2026-11-20
- Geoffrey Irving: 'Slop proofs' may stump even AI, a risk for RSI-era alignment delegation — ronbodkin · 2026-11-20
- Former Take-Two CEO Ben Feder calls GTA VI the 'last pre-AI blockbuster' — zephyr_z9 · 2026-11-20
- Tech founders surprisingly receptive to AI risk talks, says researcher — austinc3301 · 2026-11-20