Meet AlphaEvolve: The Autonomous Agent That Discovers Algorithms Better Than Humans With Google DeepMind’s Pushmeet Kohli and Matej Balog

In this episode of No Priors, Kyle Vogt interviews Pushmeet Kohli and Matej Balag, the creators of AlphaEvolve, an AI coding agent that uses Gemini models and evolutionary search to discover new algorithms. They discuss the origins of AlphaEvolve, its differences from previous projects like AlphaTensor and FunSearch, and its ability to demonstrate technical creativity. The conversation covers the mechanics of AlphaEvolve, including the use of evaluation functions and evolutionary algorithms, and its applications in optimizing data center scheduling and improving the efficiency of Google's infrastructure. They also explore the potential for self-improving AI, the implications of AI in scientific discovery and mathematics, and the collaborative role of humans in working with AI agents like AlphaEvolve. The guests also touch upon the accessibility of AlphaEvolve to the wider community.