Ryan Greenblatt corrects AI progress math: software progress needs 2.5x, not 3x, for 2x overall speedup
RyanGreenblatt · x · 2026-11-06
Ryan Greenblatt issued a correction to his earlier estimate: if 2/3 of AI progress comes from software improvements, accelerating overall AI progress 2x requires software progress to be 2.5x faster, not 3x.
The error: training compute still contributes progress at a 1/3 weight. Solving (2/3)x + 1/3 = 2 gives x = 2.5, yielding an overall estimate of 3.8x rather than 4.5x.
More from AGI Musings
- Oxford Position Paper: From Autonomous Agents to Cognitive Symbionts — Oxford · 2026-11-06
- "Give it just a few years": AGI framed as humanity's next alien-invasion moment — AndyMasley · 2026-11-06
- CDT+GT and indexical uncertainty in multi-agent clone settings: one policy network or many matters — jessi_cata · 2026-11-06
- Agents now run email and Slack for one team: a post-literacy workflow in practice — curious_vii · 2026-11-06
- Epoch AI researchers debate: inability to find slow-improving tasks is evidence of acceleration — Jsevillamol · 2026-11-06
- A Reddit deep dive with ChatGPT on whether posthuman minds could engineer suffering out of consciousness — evandegr · 2026-11-06