METR: AI capability acceleration not driven by faster training compute growth
BenBajarin · x · 2026-11-06
METR's analysis estimated how much of the acceleration in AI capabilities was due to compute, data, and algorithmic progress, concluding it was not due to an increase in the growth rate of training compute—pointing to algorithmic progress instead.
More from Infra
- Blogger Proposes Data Centers Win Locals Over With Free ChatGPT Pro and Waste Heat — flowersslop · 2026-11-06
- Voters reject data centers in 40 of 41 ballot measures; Musk eyes orbital data centers — PeterDiamandis · 2026-11-06
- Ripple-65B open-weight model runs at 7 tok/s locally on a 64GB Mac mini — DZhang50 · 2026-11-06
- Liquid AI: a 9B LLM can't run in a car, but LFMs can — JosephJacks_ · 2026-11-06
- Nvidia chips pile up in Hong Kong as China withholds import approval — pstAsiatech · 2026-11-06
- Chart: UK IT investment at 0.66% of GDP vs US 2.21%, missing the AI capex boom — KeeganMcB · 2026-11-06