Artificial intelligence learns to read numbers and models of all scientific disciplines, from mathematics to physics to astronomy, becoming an unprecedented laboratory assistant, a kind of polyglot of science, ready to assist scientists of all disciplines in analyzing the growing amount of data to help that is produced in all areas of research. The new system is called Polymathic AI and is the result of an international project that uses the same technology that underlies ChatGpt.
The initiative was launched by a group of physicists and mathematicians in a series of online articles on arXiv, the platform that hosts articles not yet subject to scrutiny by the scientific community.
Unlike ChatGpt, the new scientific AI system is not trained to work with words and sentences, but with numerical data and simulations from all areas of science, from stars to climate. “This will completely change the way people use artificial intelligence and machine learning in science,” notes Shirley Ho, principal investigator for Polymathic AI and coordinator of the Computational Astrophysics group at the Flatiron Institute in New York. “We want to make everything public,” Ho continues. “We want to democratize artificial intelligence for science so that in a few years we can provide the community with a pre-trained model that can help improve scientific analysis in a wide variety of problems and areas.”
Read the full article on ANSA.it