Chollet: intelligence is a bounded conversion ratio, not a scalar stat
sebkrier · x · 2026-11-11
François Chollet argues the biggest misconception about intelligence is treating it as an unbounded scalar like height ("future AI will have 10,000 IQ"). Intelligence, he says, is a conversion ratio with an optimality bound — improving it is less "making the tower taller" than "making the ball rounder," and past a point gains are marginal.
He notes machines will still hold advantages beyond intelligence — removing biological bottlenecks: faster processing, effectively unlimited working memory.
teortaxesTex extends this: there's a non-zero-probability timeline where coding gets solved superexponentially — <30B models in Q2 2027 zero-shot building operating systems — yet even then no RSI or singularity follows, because intelligence itself is bounded.
More from AGI Musings
- Viral post: five years of 'maybe next year,' then 2026 arrived — hey_abusiddik · 2026-11-11
- 'Chat with your data' misses the point — people want a smart coworker — vboykis · 2026-11-11
- Ben Todd: AI task time horizons appear superexponential in AECI, not just revenue — ben_j_todd · 2026-11-11
- When AI becomes an oracle, what happens to mathematics? — stevenstrogatz · 2026-11-11
- David Patterson: superintelligence will end profit and equity growth, only sales-tax-funded universal high income works — davidpattersonx · 2026-11-11
- Even after a 6-month recursive leap, AI progress stays 3x the 2024 rate — ben_j_todd · 2026-11-11