LLMs Aren't Markov Chains: Dev Pushes Back on "Next-Word Parrot" Claim
austinc3301 · x · 2026-10-04
Responding to the claim that "AI works like human minds scanning sentences and picking the most common next word," austinc3301 laid out a full rebuttal:
- The confusion stems from conflating the training task (next-word prediction) with the trained model, which does far more sophisticated things;
- Modern LLMs have extremely complex world models and mix heuristics with algorithms for different problems — a Markov chain predictor is not how they choose the next token;
- He adds that when people say AI resembles human minds, they mean higher-level emergent behavior: natural language communication, theory of mind, and modeling human values well.
The exchange highlights a common AI-literacy misconception: using the training objective to dismiss inference-time capabilities.
More from AGI Musings
- Cambridge AI safety prof David Krueger endorses critique of Anthropic's RSI strategy — DavidSKrueger · 2026-11-20
- RSI has been happening for at least 600,000 years, says economist with GDP chart — Moh1tAgarwal · 2026-11-20
- Jon Barron updates the tech sigmoid chart: the next sigmoid never arrived — songyoupeng · 2026-11-20
- "2, 4, 8, 16, do not build the death machine": ban-AI protest chant at Anthropic goes viral — JeffLadish · 2026-11-20
- Max Tegmark: RSI will find the perfect model, then the superintelligence race shifts elsewhere — rand_longevity · 2026-11-20
- Researcher crystallizes AI views: rejects 'good' outcomes that keep humans as perpetual children — wfithian · 2026-11-20