Toby Ord stands by his RL thesis: lower bits-per-FLOP ceiling explains today's jagged AI capabilities
tobyordoxford · x · 2026-09-24
Philosopher Toby Ord reaffirms his essay's thesis after being asked if he updated on the "RL doesn't have enough bits" theory: RL has a much lower ceiling on information learned per FLOP than pretraining, raising questions of why it works and how far it can go — partly answered by the bits being highly relevant. He notes his predictions held up, including RL-era AI getting much better at a subset of tasks: today's "jagged" capabilities.
More from AGI Musings
- Could Opus 5.5 'solve' children's entertainment? The hyper-specialized content idea — repligate · 2026-09-24
- Musk: Human output is ~100 bits/sec while computers hit 1 trillion, hence Neuralink — XFreeze · 2026-09-24
- Only 2 of 46 AI Rollup Firms Get an A Grade: Vista and Apollo Top the Map — curious_vii · 2026-09-24
- Dario Amodei says Anthropic will 'slow down as much as necessary' to keep AI safe — Polymarket · 2026-09-24
- AI expert Dave Linthicum calls OpenAI doomsday narrative an overreaction — DavidLinthicum · 2026-09-24
- Altman at UN: some things AI builders say sound like bad sci-fi plots — ns123abc · 2026-09-24