OpenAI AGI Economics study: half the gains from better models come from humans changing how they use them
daveholtz · x · 2026-09-18
Columbia professor daveholtz is on leave leading AGI Economics research at OpenAI. His study shows that even as models improve rapidly, those gains won't be captured automatically unless users adapt to each new iteration.
Key figure: for tasks where you steer AI toward a specific outcome, 50% of the performance improvement from a better model came from humans changing how they used it — "if we gave people DALL-E 3 instead of DALL-E 2 and said don't change anything, they would've only gotten 50% of those gains." For open-ended tasks like brainstorming, the source of gains differs. Helping users keep up with model iterations is a problem labs must take seriously.
More from AGI Musings
- Yacine: Every Enterprise Query Bakes Company Knowledge into OpenAI Weights — yacineMTB · 2026-09-18
- Roman Yampolskiy calls the AGI sprint a race to build God in podcast — MaxUnfried · 2026-09-18
- Gary Marcus: 'Rogue agents' is AI's excuse for irresponsibly built software — GaryMarcus · 2026-09-18
- yacine argues AI companies are uncontrolled super-entities eating employers — yacineMTB · 2026-09-18
- Roman Yampolskiy returns to DOAC to update his AI safety warning that reached 20M+ viewers — MaxUnfried · 2026-09-18
- Ex-OpenAI's Yacine: enterprises are the next thinking machines — learn AI or get eaten — yacineMTB · 2026-09-18