The Big Picture: A global team of scientists, with members from the University of Cambridge, is embarking on an ambitious project. They're using the technology that powers ChatGPT to create an AI tool for scientific exploration.

What's Different About This AI?

• While ChatGPT is adept at handling words, this new AI will master numbers and physics simulations.

• The goal? To assist researchers in modeling everything from massive stars to our planet's climate.

Introducing Polymathic AI: The project, named Polymathic AI, was unveiled recently. It's accompanied by a series of research papers available on the arXiv open access repository.

Why It Matters:

• Shirley Ho, the project's lead investigator, believes this will revolutionize the use of AI in science.

• The concept is akin to mastering multiple languages, making it easier to learn even more. Starting with a robust pre-trained model can be quicker and more precise than starting from zero.

Challenges & Solutions:

• Miles Cranmer from Cambridge highlighted the challenges of researching large foundation models due to immense computing power needs.

• Collaborations, like the one with the Simons Foundation, are crucial to overcoming these hurdles.

The Vision of Polymathic AI:

• It aims to reveal links between various scientific fields.

• Historically, polymath scientists could draw connections across disciplines. Polymathic AI hopes to emulate this by aggregating data from multiple fields.

Who's Involved? The project boasts a diverse team from institutions like:

• Simons Foundation

• Flatiron Institute

• New York University

• University of Cambridge

• Princeton University

• Lawrence Berkeley National Laboratory

AI in Science: A Quick Look Back

• AI tools have been used in science before, but they were often designed for specific tasks.

• Francois Lanusse points out that this siloed approach means scientists miss out on potentially valuable cross-disciplinary insights.

Polymathic AI's Unique Approach:

• It will learn from varied data sources across physics, astrophysics, and eventually even chemistry and genomics.

• The goal is to integrate various subfields into a cohesive whole.

Addressing Limitations:

• While ChatGPT has its quirks, Polymathic AI aims to sidestep these by treating numbers as numbers and using genuine scientific datasets.

A Commitment to Openness: Shirley Ho emphasizes the project's dedication to transparency. The team aspires to:

• Make all findings public.

• Democratize AI for science.

• Offer a pre-trained model to the community to enhance scientific analyses across numerous challenges and sectors.

source: Scientists begin building AI for scientific discovery using tech behind ChatGPT

Hope you enjoyed today's newsletter

Follow me on Twitter and Linkedin for more AI news and resources.

Did you know you can add Superpower Daily to your RSS feed https://rss.beehiiv.com/feeds/GcFiF2T4I5.xml

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.