Welcome to the Future of Discovery: Introducing FunSearch
Research
By searching for “functions” written in computer code, FunSearch made the first discoveries in open problems in mathematical sciences using LLMs. Large Language Models (LLMs) are useful assistants but could they discover entirely new knowledge? FunSearch introduces a method to search for new solutions in mathematics and computer science by pairing a pre-trained LLM with an automated “evaluator”. This approach resulted in the discovery of new solutions for challenging open problems and more effective algorithms.
Discovering new mathematical knowledge and algorithms in different domains is notoriously difficult, but FunSearch rose to the challenge. By addressing the cap set problem, a longstanding open problem in mathematics, FunSearch showcased its ability to surpass traditional computational solvers and generate new knowledge. FunSearch’s ability to output concise and human-interpretable programs sets it apart, making it a valuable tool for collaborative problem-solving.