AI program helps identify variables in physical concepts
To understand any physical phenomenon, one must identify the variables responsible for it. While scientists are familiar with the variables of most physical connections, some are still elusive. Now, researchers from Columbia University have used artificial intelligence (AI) to develop a program to observe such physical phenomena and detect associated variables. The program uses a video camera to observe the dynamics and then processes the information to come up with the minimal set of basic variables needed to describe it.
inside researchpublished year Natural Computational Science, the researchers started by processing raw video in a system for which they already knew the answer. They then matched the results of WHO with their own system that turned out to be close. “We think this answer is close enough. Where the work is mainly done. Especially since all AI has access to what is raw video footage without any knowledge of physics or geometry. But we want to know what the variables really are, not just their numbers.” speak Hod Lipson, director of Creative Machines Lab in the Department of Mechanical Engineering. Lipson is also the author of the study.
The team then tried to visualize the variables the program had defined. While they found two variable correspond to the angles of the arms, the other two cannot be described. “We tried correlating other variables with anything and everything we could think of: angular and linear velocities, kinetic and potential energy, and various combinations of known quantities. know. But nothing seems to match perfectly,” explains Boyuan Chen PhD ’22, assistant professor at Duke University and lead author of the study.
The researchers continued to test the system and provided videos within the system to which they had no answers. These include videos of aerial dancers and lava lights. The system gives eight variables for both. Meanwhile, for the video of the fire from the holiday fireplace loop, the system returned 24 variables.
Now the team hopes that such an AI program could help scientists decipher complex phenomena in fields ranging from biology to cosmology.