Physical Superintelligence, known as PSI, says an AI research system it developed found a proof of the Aaronson-Ambainis conjecture, a longstanding mathematical problem about when quantum computers can outperform ordinary computers.
For researchers developing AI tools for mathematics, the claim offers a test of whether an AI system can produce new results that people and software can check. For quantum-computing researchers, the result concerns the limits of a particular kind of speed advantage, rather than a new computer or a demonstration of faster hardware.
In its November 6 research announcement, PSI said its system ran on the company’s infrastructure using instances of a non-public Anthropic model. It said the system had a complete proof during the third day of the run, then improved the result twice. According to PSI, the Lean proof assistant accepted the final version.
The conjecture concerns how many times a computer must look at parts of its input to produce an answer. PSI says its proof establishes that, for most randomly chosen inputs in the setting studied, an ordinary computer can approximate a quantum computer’s answer probabilities without needing exponentially more such looks.
That is a statement about access to input data, not a general comparison of total running time. The original Aaronson-Ambainis paper describes a model that counts queries while ignoring other computational steps.
PSI has published a paper, formal proofs and verification materials. Its announcement credits David Turturean with developing and running the system, reviewing the paper’s draft and taking responsibility for its correctness. PSI says Jim Halverson audited the proof and Phiala Shanahan commented on its Lean formalization.
JBizNews Desk | Technology
© JBizNews.com. All rights reserved. Unauthorized reproduction or redistribution is strictly prohibited.
