Aarva

Noahpinion ·Future-gazing

Where’s the “intelligence Explosion”?

by Ramez Naam

Published 2026-09-27 07:20:08+00:00

If an artificial intelligence can write its own code, what stops it from improving forever?

0:00 / 45:54 · Narrator Charon

Context

Silicon Valley is currently obsessed with a concept called the fast takeoff—a sudden event where artificial intelligence learns to build smarter versions of itself in a matter of days. Writing on 27 September 2026, futurist Ramez Naam tests that anxiety against actual performance data from inside labs like OpenAI. Rather than trading in science fiction, he looks at the hard physical limits of computing power and scientific discovery. The argument asks a remarkably grounded question: what if the mechanics of research simply require too much time for the future to arrive all at once?

Show notes

An evaluation of the theory that artificial intelligence is approaching a rapid intelligence explosion. By comparing benchmark forecasts with internal productivity data from major labs, the analysis measures the actual rate of AI self-improvement. It details how diminishing returns in machine learning currently prevent a runaway acceleration to superintelligence.

Read on Noahpinion →