New Delhi – Scientists from around the world have joined forces to embark on a groundbreaking research initiative that harnesses the transformative technology underpinning ChatGPT. Aptly named “Polymathic AI,” this endeavor seeks to create an AI-powered tool for scientific discovery. While ChatGPT excels in processing words and sentences, Polymathic AI takes a different route, learning from numerical data and physics simulations spanning various scientific disciplines. Its primary mission is to assist scientists in modeling a wide array 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, shared her perspective on this revolutionary venture. She stated, “This will completely change how people use AI and machine learning in science.”
The core idea behind Polymathic AI is analogous to the concept that learning a new language becomes easier when you are already fluent in multiple languages. By commencing with a large, pre-trained foundation model, the process of creating a scientific model becomes significantly faster and more precise than starting from scratch. Surprisingly, this method can yield favorable results, even when the training data seemingly lacks a direct connection to the specific problem at hand.
One of the distinctive features of Polymathic AI is its capacity to unveil commonalities and connections between diverse fields that might have otherwise remained undiscovered. Siavash Golkar, a co-investigator and guest researcher at the Flatiron Institute’s Center for Computational Astrophysics, emphasized the potential of this interdisciplinary approach.
The Polymathic AI team comprises experts in a wide spectrum of fields, ranging from physics and astrophysics to mathematics, artificial intelligence, and neuroscience. Their project draws on data from diverse sources across the realms of physics and astrophysics. As the initiative progresses, it aims to extend its reach into other domains, such as chemistry and genomics.
While ChatGPT has been widely embraced, it does have well-documented limitations regarding accuracy. In contrast, Polymathic AI’s project is poised to circumvent many of these challenges. Shirley Ho explained that the initiative would treat numerical data as precise values, as opposed to mere characters on the same level as letters and punctuation. Furthermore, the training data will incorporate genuine scientific datasets that capture the fundamental principles governing the universe.
Transparency and openness stand as foundational pillars of the Polymathic AI project. According to Ho, “We want to make everything public. We want to democratise 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.” This commitment to openness and accessibility is poised to usher in a new era of scientific discovery with the aid of AI.