Empowering Scientific Discovery with Polymathic AI

A consortium of scientists is currently developing Polymathic AI, an innovative tool that harnesses the same technology that drives ChatGPT. In contrast to OpenAI's chatbot, which predominantly focuses on processing language, this novel model is designed to work with numerical data and physics simulations.

According to a blog post featured on the Simons Foundation website, Polymathic AI will assist scientists from diverse disciplines in modeling a wide spectrum of subjects, ranging from "supergiant stars to Earth's climate." Shirley Ho, the leader of the AI project, expressed that this new AI model is poised to revolutionize the use of AI and machine learning in scientific research.

Ho notes that while ChatGPT has recognized limitations in terms of accuracy, Polymathic AI intends to overcome these limitations by treating numerical values as actual numbers, rather than mere characters.

The team recently published a scientific paper demonstrating the remarkable capabilities of broadly pre-trained AI models. These models can effortlessly match or even surpass fine-tuned AI models tailored for complex tasks, such as simulating turbulent fluid flow. Employing large pre-trained models offers numerous advantages over building models from scratch, even when the training data is seemingly unrelated to the specific problem at hand.

Announcing:

*The Polymathic AI Initiative*

We are researching a new class of foundation models for scientific *data*, developing models that share scientific concepts across disciplines.

Site: https://t.co/C2Qsi59HzI
News: https://t.co/JomPjdIJAG
Thread: 👇 pic.twitter.com/pYvM18md4B

— PolymathicAI (@PolymathicAI) October 9, 2023

In contrast to most AI models, which are typically designed for specific use cases, the Polymathic AI project team envisions a truly cross-disciplinary model. It will draw insights from a wide array of data sources spanning physics, chemistry, genomics, and astrophysics, connecting seemingly distinct subfields into a cohesive whole that is greater than the sum of its parts.

This diverse team of scientists working on Polymathic AI hails from a range of expertise, including mathematics, artificial intelligence, neuroscience, astrophysics, and physics. They represent institutions such as the Simons Foundation, Flatiron Institute, New York University, Lawrence Berkeley National Laboratory, University of Cambridge, and Princeton University.