Experts in physics, astrophysics, mathematics, artificial intelligence, and neuroscience make up the Polymathic AI team. With the help of technologies akin to ChatGPT, Polymathic AI hopes to develop an AI-powered instrument for furthering scientific research. While ChatGPT focuses mostly on text, the recently established Polymathic AI project makes use of physics simulations and numerical data from a variety of scientific fields. This instrument will let scientists model a wide range of phenomena, from Earth’s climate to supergiant stars.
Shirley Ho, a group leader at the Flatiron Institute’s Center for Computational Astrophysics in New York City and principal investigator of Polymathic AI, claims that “this will fundamentally transform the use of AI and machine learning in scientific research.” The idea behind polymathic AI is similar to how learning a new language becomes easier when a person is multilingual. Even while the training data may not appear immediately relevant to the particular situation at hand, using a large pre-trained foundation model is typically faster and more accurate than creating a scientific model from scratch.
The Flatiron Institute’s Center for Computational Astrophysics guest researcher and co-investigator Siavash Golkar states that “polymathic AI can reveal commonalities and connections between different scientific fields that might otherwise go unnoticed.”
The Polymathic AI team aims to learn from a variety of data sources across many scientific fields, including physics, astrophysics, chemistry, and genomics. The team is composed of professionals in physics, astrophysics, mathematics, artificial intelligence, and neuroscience. Their objective is to utilize this interdisciplinary knowledge to tackle a diverse range of scientific problems.
As opposed to ChatGPT’s well-known accuracy limitations, Polymathic AI’s method treats numerical data as actual numbers instead of just letters. Additionally, actual scientific datasets that capture the fundamental physics of the universe will be included in the training data.
Openness and transparency are core values of the Polymathic AI project. Ho underlines, saying, “We’re determined to make everything available to the public. In a few years, we hope to be able to offer the scientific community pre-trained models to improve analysis across a wide range of issue domains by democratizing AI for scientific research.”