Ramez Naam's Essay Asks: Where's the "Intelligence Explosion"?
Futurist Ramez Naam, invited by economist Noah Smith, published a long essay titled "Where's the 'intelligence explosion'?" on the Noahpinion blog, laying out a skeptical case against recursive self-improvement (RSI) and the "intelligence explosion/FOOM" narrative. His conclusion: the intelligence explosion remains nowhere in sight, and RSI is hitting diminishing returns on multiple fronts.
Confirmed
- The essay, published on Noahpinion in late September, was written by Naam, author of "The Infinite Resource," known for accurately forecasting solar and battery cost declines years in advance
- Core argument: despite continuing model progress, there is a clear gap between actual AI progress and the intelligence-explosion narrative; Naam notes that classic RSI arguments typically assume AI already matches humans in every aspect of AI research, an assumption the data does not support
- Oxford scholar Toby Ord recommended the piece, citing a key data point: OpenAI's internal models currently have an 80%-time task horizon of only about 15 minutes on real research tasks, far below horizon estimates from METR (general software engineering) or UK AISI (web tasks), offering empirical support that current AI is far from triggering an explosion
Why it matters
- The intelligence explosion is a core narrative shared by both accelerationists and safety advocates; if diminishing returns to RSI hold, estimates of AI development speed, risk timelines, and policy urgency would shift significantly
- The essay was commissioned by mainstream economist Noah Smith and endorsed by Toby Ord, sparking discussion on Reddit and elsewhere—evidence that intelligence-explosion skepticism is moving from the margins into broader public debate
2026-09-27 ~ 2026-09-29 · 6 related posts
Primary sources
- Ramez Naam: The Intelligence Explosion Is Fighting Diminishing Returns, Not Happening — natolambert ·
- Ramez's new piece tests intelligence explosion claims against current AI scaling data — tobyordoxford ·
- OpenAI's Internal Models Have Only ~15-Minute Time Horizons on Real Research Tasks — tobyordoxford ·
- Ramez Naam asks: where's the 'intelligence explosion'? — JoshuaZ1 · 2026-09-27
- Futurist Ramez Naam Makes the Skeptic's Case Against Recursive Self-Improvement and the Intelligence Explosion — clarejtbirch · 2026-09-28
- [source] Ramez Naam: The Intelligence Explosion Is Fighting Diminishing Returns, Not Happening — natolambert · 2026-09-29
- [source] Ramez's new piece tests intelligence explosion claims against current AI scaling data — tobyordoxford · 2026-09-29
- [source] OpenAI's Internal Models Have Only ~15-Minute Time Horizons on Real Research Tasks — tobyordoxford · 2026-09-29
1 near-duplicate retellings: ccerrato147