METR Says Anthropic Models Are Improving Twice as Fast as a Year Ago

According to ChainCatcher citing The Verge, AI safety research organization METR said after an 11-week study of Anthropic that the company’s models are improving about twice as fast as they were a year ago. METR said progress over the past year was equivalent to roughly two years of earlier advances. As of October, AI was responsible for most of the acceleration in research and development, and its contribution was growing. The acceleration mainly comes from AI assisting human researchers, rather than working autonomously. In an informal survey, Anthropic employees using the company’s most capable internal models self-reported average productivity about 12 times higher than without AI assistance. Anthropic heavily redacted its public report. Co-founder Sam McCandlish said the information involved commercially sensitive material and could help other companies accelerate their research and development.