A recent MIT graduate working at startup lab Physical Superintelligence (PSI) has used early access to Anthropic's Fable 5.2 model to solve an 18-year-old quantum computing problem, the Aaronson-Ambainis conjecture. David Turturean deferred a master's at MIT's computer science and AI lab to join PSI, and told TechCrunch he accepts the result might be his last paper. "In a few months, like, why would you even put out a paper?" he said.
The conjecture concerns whether quantum computers can solve certain problems faster than classical ones. Scott Aaronson, one of the researchers who posed it, wrote that he doesn't fully understand the proof but believes it is correct because it was autoformalized in Lean, a coding language for math, and explained by top researchers PSI shared it with. The solution is a negative result: it closes off one avenue for quantum speedups, not all of them. "We're no longer looking there," Turturean said.
Aaronson described feeling satisfied, awestruck, disappointed and fatigued, and said he was surprised the problem was missing from OpenAI's recent math results. PSI CEO Matt Pines said effective research systems need harnesses, tools, memory systems and simulation infrastructure to pursue problems at breadth and depth.
PSI currently earns revenue by optimizing data centers, simulating heat, power, water and sound variables. The longer-term goal is turning theoretical physics breakthroughs into technologies. Pines said that six to nine months ago he had to stretch to convince investors that theoretical physics would soon be mostly the domain of machines. "It's not there yet, but it's on its way," he said.
Pines noted that physics differs from math because "nature gets a vote in your models of reality," and said PSI focuses on validating computational results through experiments. Jim Halverson, a Northeastern University professor who advises PSI, said best practices are still emerging and that the company chose transparency and contacting experts. Turturean said reading the AI-generated solution made him realize "the weight of what AI can do."
Source: TechCrunch