Why physicists flocked to ML: reductionism dead-ended, and models became the tool for irreducible complexity

MickeySteamboat · x · 2026-09-24

Reacting to tszzl's observation that many physics-trained researchers—John Schulman, Liam Fedus, and Anthropic's Dario Amodei and Jared Kaplan among them—concluded ML would produce more scientific progress than physics, beffjezos adds his own take: the reductionist approach to physics was a dead end, and ML is the ultimate tool for pockets of irreducible complexity—"programmable complexity to mirror the complexity of the universe."

Related event: Physicists Flock to AI, Predicting Breakthroughs in Mathematical Physics(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →