AI & agents

Claude improves best-known quantum circuit designs

Published
Source
Anthropic Research

Summary

Anthropic reports that it gave Claude plain-language specifications for eight quantum circuits spanning chemistry, optimization, physics simulation, and number theory, plus an automated evaluator that scored each design. Working in a closed loop without human input between rounds, Claude produced designs using fewer qubits and gates than the best previously published circuits, in some cases by an order of magnitude or more, according to the company. The work was done with Harvard collaborators and the paper is on arXiv.

Why it matters for our work

With numerically verifiable objectives and automated scoring, an AI agent can iteratively improve specialized research designs on its own. Human researchers shift toward defining problems and validation criteria rather than hand-optimizing designs.

Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.

Read original (opens in a new tab)