Scientists Collaborate to Create AI Tool for Scientific Discovery
New York, October 15: An international team of scientists has initiated a groundbreaking research collaboration to harness the technology underpinning ChatGPT, aiming to develop an AI-powered tool designed to advance scientific discovery.
Distinct from ChatGPT’s language-oriented capabilities, this endeavor, named Polymathic AI, will specialize in learning from numerical data and physics simulations spanning various scientific domains. Its objective is to assist scientists in modeling a wide spectrum of phenomena, from supergiant stars to the Earth’s climate.
Shirley Ho, the principal investigator of Polymathic AI and a group leader at the Flatiron Institute’s Center for Computational Astrophysics in New York City, emphasized the transformative potential of this initiative, stating, “This will completely change how people use AI and machine learning in science.”
The concept behind Polymathic AI is likened to language acquisition; starting with a large, pre-trained foundation model can significantly expedite and enhance the creation of scientific models, even when the training data seems unrelated to the specific problem at hand. Siavash Golkar, a co-investigator from the Flatiron Institute’s Center for Computational Astrophysics, highlighted the tool’s potential for identifying connections between different scientific domains.
Polymathic AI comprises experts in physics, astrophysics, mathematics, artificial intelligence, and neuroscience. Its project will involve learning from diverse datasets across physics and astrophysics, with plans to extend its applications into fields like chemistry and genomics.
While ChatGPT is known for certain limitations related to accuracy, Polymathic AI aims to overcome these challenges by treating numerical data as genuine values, not mere characters, and by utilizing authentic scientific datasets that encapsulate the intricacies of the cosmos.
Transparency and open access are integral to the project, as Shirley Ho affirmed, “We want to make everything public. We want to democratize AI for science in such a way that, in a few years, we’ll be able to serve a pre-trained model to the community that can help improve scientific analyses across a wide variety of problems and domains.”
