Researchers in the UW—Madison American Family Insurance Data Science Institute (DSI) are among three team projects recently awarded funding as part of a $6.4 million investment from the U.S. Department of Energy (DOE) for research in artificial intelligence for high energy physics. These projects are part of a DOE Office of Science initiative supporting research that pushes beyond what is currently possible in high energy physics, advancing novel AI applications that enable new lines of investigation.
“These awards represent new partnerships between researchers at DOE National Labs, universities, and the private sector that will enable the next discoveries in high energy physics,” says Harriet Kung, DOE Acting Associate Director of Science for High Energy Physics.
Kyle Cranmer, Professor of Physics and Director of the American Family Insurance Data Science Institute (DSI) leads the UW—Madison effort in collaboration with Lance J. Dixon of the SLAC National Accelerator Laboratory, operated by Stanford University. Their project will leverage recent advances in transformers, a type of AI technology that has led to recent breakthroughs in natural language processing. However, instead of working with written languages like English, these AI systems will learn to speak the language of mathematics that describes how fundamental particles interact. The goal is to discover hidden patterns and equations in what physicists call “scattering amplitudes.”
“This is a very compelling cross-pollination of ideas. By adapting AI techniques developed for language, we have a much more straightforward path to discovering patterns that will be interpretable by humans,” says Cranmer.
The UW—Madison and Stanford groups will also collaborate with François Charton at Meta AI in Paris and Matthias Wilhelm at the Niels Bohr Institute in Copenhagen.
“We have an awesome team,” says Cranmer. “Lance and Mattias are among the world’s leading experts in scattering amplitudes, François has had enormous success in using transformers for symbolic mathematics, and I am able to translate and facilitate, having some level of expertise and intuition for both the AI and physics aspects of the problem.”
This award will support postdoctoral researchers at both UW-Madison and SLAC.