Google DeepMind’s new AI agent uses large language models to crack real-world problems
1 min read
Summary
London-based AI company DeepMind, owned by Alphabet, the parent company of Google, has developed a coding agent called AlphaEvolve that generates algorithms to solve complex problems using language models.
While previous versions of the system focused on games, AlphaEvolve uses large language models such as Gemini 2.
0 Flash and Gemini 2.
0 Pro to generate many potential solutions to a given problem, iterating and combining them to produce the best algorithm.
AlphaEvolve has already improved Google’s data centre operations and chip performance.
The new tool can be applied to any problem that can be described in code and has solutions that can be evaluated by a computer.
However, it currently cannot be used on problems that need human scoring.