Christiano: Two Months of Math Progress in a Decade, and ML Could Be Next
OwainEvans_UK · x · 2026-11-21
A relay of Paul Christiano's arguments on the Dwarkesh podcast, framed as three points:
- In two months, the field has just seen 10 years of progress in maths.
- The same could soon happen in machine learning itself.
- That would likely produce AI agents able to learn any job to a superhuman level.
Christiano also lays out a concrete scenario: on the current trajectory, absent any slowdown, there is roughly a 1-in-3 chance of superintelligence powerful enough to overthrow the government within a year. He argues labs should pause further algorithmic progress until they can secure their weights and research secrets — otherwise China could steal the superintelligence mid-explosion, leaving labs unable to slow down even if they see models going off the rails.
Related event: Christiano Warns AI Takeover Risk Could Emerge Within 6-12 Months(2 posts)→
More from AGI Musings
- AI safety researcher: doubting China's moratorium compliance is like doubting your own marriage — DavidSKrueger · 2026-11-21
- Two Riemann hypothesis proofs announced same day: one with Claude, one "all human, no AI" — Dr_Atoosa · 2026-11-21
- Hinton may be softening on an AI moratorium after SF risk-scenarios panel — DavidSKrueger · 2026-11-21
- Thinking Machines releases Inkling 2 Mini, a looped transformer topping sub-100B open-weights coding benchmarks — SonglinYang4 · 2026-11-21
- Mathematicians will study Lean proofs like N-body simulation results, researcher predicts — mjuric · 2026-11-21
- 205 Protesters Rally Outside Anthropic Against Recursive Self-Improvement — DavidSKrueger · 2026-11-21