Noah Smith
· Noahpinion
· March 22, 2026 at 14:57
· ⏱ 74 min read
| Read on Substack ↗
Summary
Noah Smith argues that AI will revolutionize science not by replacing human creativity but by exploiting complex, incompressible regularities (Cloud Laws) that humans cannot intuit or communicate. This shifts the frontier from fundamental physics (mostly mapped) to biology, materials, and social systems, where vast combinatorial spaces remain untapped. For markets, the implication is that AI-driven discovery will accelerate progress in applied fields like materials science and drug development, but with long commercialization timelines and no direct investment signals from the author.
•Claude estimates 25-35% probability that AI produces a novel conceptual framework in some scientific field within 5 years.
•Noah Smith predicts AI will discover non-compressible scientific patterns that humans could never have found, analogous to how LLMs learned language without explicit rules.
•Dhonz, a materials scientist, calls most AI-driven materials timelines 'pure fantasy' and emphasizes that current LLMs are stateless systems without causal understanding.
•The article identifies topological materials as a prime example where AI can fill the incompressible mapping between chemistry and topology.
•Claude argues that AI's biggest leverage is in fields with vast combinatorial search spaces, weak theoretical guidance, and short experimental feedback loops (e.g., materials, drug discovery).
•Noah Smith uses Lagrange's quote to suggest fundamental physics may be fully discovered, leaving only technologically inaccessible regimes.
•The conversation concludes that the ultimate payoff of AI may be improving human experience (health, psychology, society) rather than conquering the physical universe.
•The article references Q from Star Trek: TNG to frame the shift from mapping stars to charting 'the unknowable possibilities of existence.'