MIT researchers say AI may soon dominate theoretical science

By Splitfeed EditorialLast updated Oct 14, 2026, 19:00 UTC

On October 14, 2026, MIT researchers told WIRED that artificial intelligence may soon become the default method for solving theoretical problems across multiple scientific disciplines, with economics and high-energy physics named as specific targets. The claim rests on what the researchers describe as measurable progress AI has already made in mathematics, which they are treating as a proof of concept for extending the approach into other domains. No specific model, system, or quantitative benchmark was cited in the available material.

The framing matters because calling something a "default method" is a strong claim in academic culture, where methodological consensus shifts slowly and is usually contested. Theoretical work in fields like high-energy physics and economics involves not just computation but the construction and interpretation of formal arguments, the kind of work that has historically resisted automation even as other parts of research have been assisted by it. If the mathematics analogy holds, the implications extend well beyond convenience tools and into questions about who or what drives intellectual progress in science.

With no published reactions yet from the Pro-AI, Anti-AI, or Middle Ground camps, the likely positions can be sketched from typical patterns. Enthusiasts would probably point to the mathematics precedent as evidence that the skeptics who once doubted AI in that domain were wrong, and argue the same pattern will repeat. Critics would likely challenge the leap from mathematics, a domain with clear formal rules and verifiable proofs, to fields where problem definitions and interpretive frameworks are themselves contested. A middle ground position would probably welcome AI as an accelerant for certain kinds of theoretical work while questioning whether "default method" overstates the case and pushing for specifics about what tasks, exactly, are being displaced.

The argument will sharpen considerably if and when MIT researchers or others publish concrete results showing AI solving open theoretical problems in economics or high-energy physics at a level peer reviewers accept as genuine contributions. That kind of empirical demonstration, rather than a general forecast, would give all three camps something concrete to argue over.