Anthropic AI Proves 2011 Quantum Conjecture, Narrowing Speedups for Unstructured Problems
Updated
Updated · New Scientist · Nov 12
Anthropic AI Proves 2011 Quantum Conjecture, Narrowing Speedups for Unstructured Problems
1 articles · Updated · New Scientist · Nov 12
Summary
Fable 5.2 produced a proof of the 2011 Aaronson-Ambainis conjecture, showing quantum computers do not gain a major edge on a class of random yes-or-no problems.
4 days of work by the model yielded a proof that PSI said it refined twice and verified in Lean before outside mathematicians at Stanford wrote an exposition of its key ideas.
That result removes one candidate route to quantum advantage, though Scott Aaronson and Shachar Lovett said the practical impact is limited because the ruled-out direction was not the likeliest source of useful speedups.
Stanford researchers said the proof still leaves many openings for quantum gains, including shorter or less random inputs, non-binary outputs, molecule simulation and some optimization problems.
The breakthrough adds to a growing pattern of AI settling well-posed questions in quantum information science, pushing researchers toward framing harder, more creative problems.