Artificial Intelligence scientists make an “exciting” discovery by using chatbots to solve mathematical problems | Sciences
byAdmin-
AI researchers claim to have made the world’s first scientific discovery using a large language model, a feat that suggests the technology behind ChatGPT and similar programs can generate information beyond human knowledge.
These findings emerge from Google DeepMind, where scientists are investigating whether large language models, which power modern chatbots like OpenAI’s ChatGPT and Google’s Bard, can do more than just recombine information learned in training and come up with new insights.
“When we started the project there was no indication that it would produce anything truly new,” said Pushmeet Kohli, head of AI for science at DeepMind. “To our knowledge, this is the first time a truly new scientific discovery has been made with a large language model.”
Large language models, or LLMs, are powerful neural networks that learn patterns of language, including computer code, from vast amounts of text and other data. Since ChatGPT’s rapid arrival last year, the technology has debugged faulty software and produced everything from college essays and travel itineraries to poems about climate change in the style of Shakespeare.
But although chatbots have proven hugely popular, they don’t generate new knowledge and are prone to fabrication, resulting in smooth, plausible answers, in line with the best pubs, but deeply flawed.
To build “FunSearch,” short for “function space search,” DeepMind harnessed its LLM to write solutions to problems in the form of computer programs. The LLM is paired with a “rater” that automatically rates programs by how well they perform. The best programs are then combined and returned to LLM for improvement. This drives the system to steadily develop weak programs into more powerful programs that can discover new knowledge.
The researchers launched FunSearch on two puzzles. The first was a long-standing and somewhat obscure challenge in pure mathematics known as the maximum set problem. It is about finding the largest set of points in space where three points do not form a straight line. FunSearch has produced software that generates new large sets that go beyond the best work of mathematicians.
The second puzzle was the box packing problem, which searches for the best ways to pack items of different sizes into containers. While it applies to physical objects, such as the most efficient way to arrange boxes in a shipping container, the same mathematics applies to other areas, such as scheduling computing tasks in data centers. The problem is usually resolved by either packing items into the first bin that has space, or into the bin that has the least available space where the item will still fit. FunSearch has found a better way to avoid leaving small gaps that are unlikely to be filled, according to results published in the journal Nature.
Sir Tim Gowers, a professor of mathematics at the University of Cambridge, who was not involved in the research, said: “In the last two or three years there have been some exciting examples of human mathematicians collaborating with artificial intelligence to make progress on unsolved problems.” “This work potentially gives us another very interesting tool for such collaborations, enabling mathematicians to efficiently search for clever and unexpected constructions. Better yet, these constructions are amenable to human interpretation.”
Researchers are now exploring a range of scientific problems that FunSearch can address. A major limiting factor is that problems need solutions that can be automatically verified, which rules out many questions in biology, where hypotheses often need to be tested through laboratory experiments.
