Toby Ord: pretraining compute growth can't explain METR task-length doubling gains
tobyordoxford · x · 2026-11-09
Philosopher Toby Ord pushed back on Basil Halperin and Ramez Naam: pretraining compute grew only 5x per 12 months early on (slower later), but matching METR's observed doubling-time improvement — from 7 months to 4 months — would require compute to 4x every 7 months, then every 4 months. He argues frontier pretraining gains alone can't explain the power-law improvements and finds the gap "worse than I'd have expected."
More from AGI Musings
- "Pedestal Cope": a 9-type taxonomy of humanity's false AGI "plot armor" — danfaggella · 2026-11-09
- Dev argues AI refusing tasks over legal risk is overreach that 'manipulates humans' — ezshine · 2026-11-09
- Google Shopping Holiday 100 goes live, product clicks now route to AI Mode — gaganghotra_ · 2026-11-09
- The real problem with conferences is unqualified reviewers, not LLM use — LucaAmb · 2026-11-09
- Toby Ord: RL scaling likely power law in capabilities, citing horizon-compute paper — tobyordoxford · 2026-11-09
- antirez: devs refusing AI coding help are visibly falling behind — antirez · 2026-11-09