AI's Math Takeover Is Spreading to Physics and Economics
Original: AI Came for Math. Now It’s Chasing the Rest of Science
Why This Matters
Rapid AI gains in theoretical science could reshape how entire research fields operate.
MIT researchers found AI models jumped from solving ~2% to ~35% of theoretical problems in physics, economics, and statistics between mid-2024 and mid-2026, signaling a wave similar to math's ongoing AI disruption.
At a recent MIT talk, researcher Neil Thompson presented findings that suggest AI is moving well beyond mathematics into fields like probability, statistics theory, economic theory, high-energy physics, gravitation, cosmology, and quantum physics. The study—co-authored with PhD student Gabriel Manso and research scientist Claudio Mayrink Verdun—tested OpenAI models including o1, o3, GPT-5, GPT-5.1, GPT-5.4, and GPT-5.5 against real research problems drawn from papers posted between December 2025 and May 2026. Critically, the solutions came from papers published after each model's training cutoff, so the models had to reason independently rather than recall memorized answers. Internet access was also blocked during testing.
The jump from 2% to 35% correct solutions in two years, with no apparent slowdown, led Thompson to conclude: "Within just a few years, we expect a lot of theory questions to be totally transformed in terms of AI's ability to do them." The study focused specifically on fields where conclusions can be reached through reasoning alone, without simulations or physical experiments.
This comes as math is already in turmoil: OpenAI recently announced an unreleased model solved the long-standing Navier-Stokes challenge, sparking a priority dispute with an NYU mathematician who alleged OpenAI acted after learning of competing work involving an Anthropic-affiliated researcher. Hundreds of additional math breakthroughs have followed, prompting calls from mathematicians for limits on AI use in the field.