Get the latest tech news
Google DeepMind’s AI Agent Dreams Up Algorithms Beyond Human Expertise
A new system that combines Gemini’s coding abilities with an evolutionary approach improves datacenter scheduling, chip design, and fine-tune large language models.
A new project from Google DeepMind shows that with a few clever tweaks these models can at least surpass human expertise designing certain types of algorithms—including ones that are useful for advancing AI itself. Matej Balog, one of the research leads on AlphaEvolve, says that it is often difficult to know if a large language model has come up with a truly novel piece of writing or code, but it is possible to show that no person has come up with a better solution to certain problems. Josh Alman, an assistant professor at Columbia University who works on algorithm design, says that AlphaEvolve does appear to be generating novel ideas rather than remixing stuff it’s learned during training.
Or read this on Wired
