AI Could Transform Probability, Physics and Cosmology Within Years, MIT Research Suggests
Several theoretical scientific fields could be dramatically reshaped by artificial intelligence in the next few years, according to research presented by MIT computing and AI scholar Neil Thompson.
At a recent MIT talk, Thompson argued that the rapid progress of frontier AI models could upend probability, statistical theory, economic theory, high-energy theory, gravity and cosmology, and quantum physics.
AI’s Ability to Solve Theoretical Problems Is Rising Rapidly
Thompson’s latest research suggests that AI’s impact may soon extend beyond mathematics. According to the study, between mid-2024 and mid-2026, the best AI models improved from correctly solving about 2 percent of theoretical problems in these fields to solving approximately 35 percent.
The researchers found no sign that this progress is slowing down.
“We expect that in just a few years, many theoretical questions will be completely transformed in terms of the capabilities of AI,” Thompson says.
How the AI Research Was Conducted
Thompson’s lab examines how advances in computer science affect society and influence innovation. The study was conducted by Thompson with doctoral student Gabriel Manso and research scientist Claudio Meyrink Verdun.
The researchers selected problems described in research papers posted online between December 2025 and May 2026. They then tested whether several OpenAI models—including o1, o3, GPT-5, GPT-5.1, GPT-5.4 and GPT-5.5—could solve them.
Importantly, the solutions to these problems appeared in papers published after the models had been trained. This meant the AI systems had to develop solutions independently rather than rely on answers included in their training data. The researchers also blocked internet access, preventing the models from searching for solutions online.
Why Theoretical Science Could Be Next
The research covered a limited range of scientific fields and focused on areas where AI models could understand problems without running simulations or conducting experiments. Even so, the findings suggest that the shockwaves currently being felt in mathematics could spread to other areas of scientific inquiry where mathematical reasoning and theory are central.
OpenAI’s Math Breakthroughs Have Sparked Debate
Last month, OpenAI announced that an unreleased model had solved a problem involving Navier–Stokes equations, which describe the behavior of fluid motion.
The episode became even more controversial after a New York University mathematician who had been working with another academic associated with Anthropic accused OpenAI of monitoring them after learning of their progress.
OpenAI and other organizations have since published hundreds of solutions to long-standing mathematical problems, contributing to a crisis of sorts in the field. Some mathematicians have called for restrictions on the use of AI, warning that the process of solving a problem can be just as valuable as the final answer.
Source: www.wired.com