Two ways software progress in pretraining counts as evidence that AI task speed is accelerating
datagenproc · x · 2026-11-12
datagenproc frames two interpretations for pretraining-software, inference-efficiency, or data-efficiency gains: (1) they are intermediate factors in the (compute, data, labor) → AECI production function, so acceleration hints future acceleration but is hard to update on without knowing the function; (2) they are themselves well-specified tasks, so their rapid gains are early evidence of how fast such tasks can get when seriously attempted.
More from AGI Musings
- Amateurs with Claude + Codex claim discovery of a new dwarf planet in public telescope data — willdepue · 2026-11-12
- Adam Scholl: strong evidence AI is useful, weak evidence it generalizes — gleech · 2026-11-12
- "SF is right long-term, NYC right near-term, nobody has a clue mid-term" on AI — austinc3301 · 2026-11-12
- Andrew Critch: stop calling extinction risks "improvement"—RSM vs RSI matters — AndrewCritchPhD · 2026-11-12
- Canada tech vacancies down 46.9% from peak; software engineer roles fell 72% — RichardsonDx · 2026-11-12
- Any US State Could Legalize AI Safety Coordination via Parker Immunity — TinfoilTricorn · 2026-11-12