“Artificial intelligence researchers claim to have made the world’s first scientific discovery using a large language model, a breakthrough that suggests the technology behind ChatGPT and similar programs can generate information that goes beyond human knowledge.”
– Ian Sample
Google’s DeepMind has found a new way of using large language models (LLMs) that applies their processing to improve their ability. DeepMind is the AI group within Google and one of the global leaders in AI research.
The breakthrough solves complex problems by having the LLM generate two algorithms and then having them evolve together; one creates a solution and the other evaluates how valuable that solution may be.
The iterative process repeats itself until the optimal solution is proposed. The researchers call their new iterative algorithm Funsearch.
DeepMind says Funsearch has solved two complex math problems. They say these new capabilities are limited to problems that can be proven by math instead of experimentation.
AI scientists make ‘exciting’ discovery using chatbots to solve maths problems | THE GUARDIAN | December 14, 2023 | by Ian Semple