Google DeepMind's new AI agent, AlphaEvolve, represents a significant advancement in using AI to solve complex math and science problems. Built upon Gemini LLMs, AlphaEvolve employs an evolutionary approach, evaluating and refining algorithms across various applications. Unlike Gemini chatbots, AlphaEvolve minimizes inaccuracies through an automatic evaluation system, generating and improving solutions iteratively. Its general-purpose nature allows application across diverse research fields, with Google already reporting successful internal deployments.

Prepared by Jonathan Pierce and reviewed by editorial team.