What changed
David Turturean, a recent MIT graduate and technical staff member at Physical Superintelligence (PSI), used early access to Anthropic’s Fable 5.2 model to solve the Aaronson-Ambainis conjecture, an 18-year-old problem in quantum computing. The solution, which closes one avenue for quantum computers to solve certain problems faster than classical computers, was autoformalized in Lean, a coding language for math, and explained by a team of top researchers with whom PSI shared the solution.
What this means in practice
The achievement reflects a broader shift in scientific research driven by AI. Scott Aaronson, one of the researchers who posed the problem, noted that the solution was one of the last remaining open problems in quantum query complexity. PSI’s approach involves specialized model harnesses and scaffolding to solve scientific problems, aiming to create an effective research system that allows for breadth and depth in problem-solving. The company currently generates revenue by optimizing data centers using its tools, simulating variables like heat, power, water, and sound. The long-term goal is to convert theoretical physics breakthroughs into actual technologies.
Limitations and availability
The solution appears to close one avenue for quantum computers but not all of them. While AI has made significant strides in fields like mathematics and computer science, applying it to less abstracted disciplines like the life sciences or physics remains challenging. PSI is focused on validating computational results with experiments efficiently. The company had to hold its result until Anthropic’s model was nearly ready for release, allowing them to share it with relevant researchers. The academic community is still figuring out standards for this new approach to science, with transparency and expert consultation being key priorities.
