Anthropic’s Claude Mythos 5 model autonomously designed quantum circuits that outperformed previous state-of-the-art designs on seven out of eight benchmark tasks. In the most notable case, a modular inversion architecture achieved up to 17 times fewer Toffoli gates than previous approaches. The project was led by researcher Andi Gu through Anthropic’s STEM Fellows Program, in collaboration with the Harvard Quantum Initiative and QuEra Computing. The approach uses a closed-loop evolutionary framework where Claude proposes circuits, an automated judge evaluates them, and the model refines its proposals based on feedback. These results illustrate AI’s potential to contribute to scientific research by optimizing quantum circuit design with minimal human intervention.
Source: Read the original article
