Taming chaos with physics and AI
In many situations, chaos makes it nearly impossible to predict what will happen next. Nowhere is this more apparent than in weather forecasts, which are notorious for their unreliability. But the clever application of artificial intelligence can help reign in some chaotic systems, making them more predictable than ever before.
In this episode of Relatively Certain, Dina sits down with Michelle Girvan, a physics professor at the University of Maryland (UMD), to talk about how artificial intelligence can help predict chaotic behavior, as well as how combining machine learning with conventional physics models might yield even better predictions and insights into both methods.
Girvan collaborated with several colleagues at UMD on these chaos-taming projects, including physics professor Edward Ott, mathematics professor Brian Hunt, physics postdoctoral researcher Zhixin Lu, physics graduate students Jaideep Pathak and Sarthak Chandra, and physics undergraduate students Alexander Wikner and Rebeckah Fusol.
This episode of Relatively Certain was produced by Dina Genkina, Chris Cesare and Emily Edwards. It features music by Dave Depper, David Hilowitz, Blue Dot Sessions and Scanglobe. "Lorenz Attractor" is used courtesy of Michelle Wilber. Prints are available for purchase at FineArtAmerica.com. Relatively Certain is a production of the Joint Quantum Institute, a research partnership between the University of Maryland and the National Institute of Standards and Technology, and you can find it on iTunes, Google Play or Soundcloud.
Recent Podcast Episodes
Topology—the mathematical study of shapes that describes how a donut differs from a donut hole—has turned out to be remarkably relevant to understanding our physical world.
Software just might be the unsung hero of physics labs.
Chaos. Time travel. Quantum entanglement. Each may play a role in figuring out whether black holes are the universe’s ultimate information scramblers.