Jason Wei's Stanford talk: intelligence is becoming a commodity as adaptive compute takes off
dotey · x · 2026-09-18
Jason Wei, now at Meta Superintelligence Labs after co-creating OpenAI's o1 and Deep Research, gave a 30-minute talk at Stanford AI Club on frameworks for understanding AI in 2025:
Intelligence is becoming a commodity
- AI progress has two phases: pushing the frontier, then rapid commoditization once a capability lands. MMLU data shows the cost of reaching a given intelligence level keeps falling year over year.
- The key driver is that adaptive computation finally works: since o1, models dynamically allocate inference compute based on problem difficulty, driving down costs on easy tasks.
- Information access time is shrinking exponentially: o3 can't answer "how many couples married in Busan in 1983?" but OpenAI's Operator can, by navigating Korea's KOSIS database. OpenAI's BrowseComp benchmark measures such tasks — humans average 2+ hours per question, while Deep Research solves about half.
- Implications: knowledge moats get democratized (vibe coding, personal health management); private/internal information gains relative value as public info becomes free; everyone will get a personalized internet.
Related event: AI Researchers Debate Human Agency as Intelligence Becomes a Commodity(4 posts)→
More from AGI Musings
- François Chollet: modeling brain wiring as an engineering problem constrained by genome size and time — CatAstro_Piyush · 2026-11-04
- Most US Couples Now Meet in a Data Centre, Doubling Office Matches — menhguin · 2026-11-04
- Sam Altman teases next goal: 'closer to this milestone than I expected' — ChrisGPT · 2026-11-04
- Indie researcher open-sources 10 years of math work: 146k files, AI-accelerated over 3 years — HamOnWhy · 2026-11-04
- RLHF's scalar reward hides a 2,450-year-old problem: helpfulness and harmlessness are opposing forces — vasilisvj · 2026-11-04
- Gary Marcus: 'fleecing the public with AI fantasies just got harder' as AI stocks drop — GaryMarcus · 2026-11-04