To prepare AlphaGeometry’s language mannequin, the researchers needed to create their very own coaching information to compensate for the shortage of current geometric information. They generated practically half a billion random geometric diagrams and fed them to the symbolic engine. This engine analyzed every diagram and produced statements about their properties. These statements have been organized into 100 million artificial proofs to coach the language mannequin.
Roman Yampolskiy, an affiliate professor of pc science and engineering on the University of Louisville who was not concerned within the analysis, says that AlphaGeometry’s means exhibits a major development towards extra “sophisticated, human-like problem-solving skills in machines.”
“Beyond mathematics, its implications span across fields that rely on geometric problem-solving, such as computer vision, architecture, and even theoretical physics,” mentioned Yampoliskiy in an e-mail.
However, there’s room for enchancment. While AlphaGeometry can solve problems present in “elementary” arithmetic, it stays unable to grapple with the kinds of superior, summary problems taught at college.
“Mathematicians would be really interested if AI can solve problems that are posed in research mathematics, perhaps by having new mathematical insights,” mentioned van Doorn.
Wang says the aim is to use an analogous method to broader math fields. “Geometry is just an example for us to demonstrate that we are on the verge of AI being able to do deep reasoning,” he says.