AI training efficiency could make the same capability 1,000x cheaper in 10 years
natolambert · x · 2026-07-24
The argument here is that the right analogue to Moore’s law for AI is not scaling laws, but the year-over-year gains in intelligence efficiency from using models to help train models.
If training efficiency improves 2x every year, then a fixed performance level would become 32x cheaper in 5 years and 1,000x cheaper in 10 years. The implication is that intelligence could eventually behave like a commodity, similar to electricity, even if scaling the biggest models keeps helping.
Related event: AI Intelligence Efficiency Predicted to Follow Moore's Law(2 posts)→
More from AGI Musings
- Goldman finds statistically significant negative link between AI adoption and US employment growth — ben_j_todd · 2026-10-27
- a16z: World-changing tech makes a chaotic entrance — AI agents are following the same path — a16z · 2026-10-27
- NEJM AI on medical education: mastery requires performing without the machine — BraydonDymm · 2026-10-27
- Dev warns abliterated models could wreck software supply chains with 1000+ transitive dependencies — nitarshan · 2026-10-27
- Delete your priors: why every SOTA model drop should reset how you think the world works — zack_overflow · 2026-10-27
- Replit predicts every Fortune 500 will be a 'Self-Driving Company' by end of 2030 — amasad · 2026-10-27