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
- Economist Paul Novosad: writing classes shouldn't encourage AI use — paulnovosad · 2026-11-11
- Bot Gaffe tracks AI disasters, from Claude Code deleting a Mac's Library folder — SuB8u · 2026-11-11
- The human edge AI still lacks: heresy — _akpiper · 2026-11-11
- Harvard economist Deming: use AI to polish writing and you lose the learning — paulnovosad · 2026-11-11
- Tyler Cowen's conversation with Terence Tao is out, covering AI and beyond — tobias_rees · 2026-11-11
- Elvis Saravia: agents will subsume traditional software — and can even cut code bloat — omarsar0 · 2026-11-11