This AI Rejects Your Physics and Replaces it With its Own This AI Rejects Your Physics and Replaces it With its Own
One of the main benefits of artificial intelligence is how AI is able to complement human labor and enhance our abilities,... This AI Rejects Your Physics and Replaces it With its Own

One of the main benefits of artificial intelligence is how AI is able to complement human labor and enhance our abilities, both in physical tasks and cognitively demanding roles. Now, thanks to a study by Columbia University, an AI program has observed physical phenomena and likely revealed new alternative ways to describe physics and the universe.

Physics is described using variables, three of the best known, energy, mass, and velocity, that are used by scientists to paint a picture of the universe. But, if seen from a different view, can other variables be discovered and also be used to describe the same phenomena? In the study, the team fed their AI program raw video of different physics phenomena. For this phase, they already knew the solutions.

They wanted to attempt to find a minimal set of fundamental variables to describe the observed dynamics. The video used was one of a double-pendulum swinging due to be known to have four state variables. Both arms and their angle and angular velocities. After hours of being fed a video of the pendulum in use, the AI program provided the answer of 4.7 variables.

Director of the Creative Machines Lab in the Department of Mechanical Engineering said, “ We thought this answer was close enough…Especially since all the AI had access to was raw video footage, without any knowledge of physics or geometry. But we wanted to know what the variables actually were, not just their number.”

With that, the team decided to move forward with more visualizations for the program to identify the variables. There was a problem. The program was unable to describe to the team these variables in any induvial that would be understandable to humans. But, after some testing and investigation, it seemed that the two variables the program chose almost correspond with the arms. Though the other two are still a mystery. 

Boyuan Chen, assistant professor at Duke, who led the work stated “We tried correlating the other variables with anything and everything we could think of: angular and linear velocities, kinetic and potential energy, and various combinations of known quantities,…But nothing seemed to match perfectly.” The team was confident that the AI had found a valid set of four variables, since it was making good predictions, “but we don’t yet understand the mathematical language it is speaking.”

It was time to expand and feed the AI program even more visual data. So the team provided the program with a video of an air dancing in front of a used car lot. After serval hours of analyzing the visual data, it outputted eight variables. Then, they provided a video clip of flames from a holiday fireplace loop, as seen on Netflix and YouTube. The program returned twenty-four variables.

During each experiment. The number of variables was the same each time the Aritifal Intelligence program restarted, but specific variables kept changing each time. In short, there seems to be merit in the idea of alternative methods of describing the universe. Which could lead to us questioning how we do so currently. If an AI sees what human sees and is able to describe variables in a different manner, one has to ask themselves if aliens perceived the universe in the manner we see. Though the universal language of mathematics, as are as we know it, is universal, could alternative methods of describing physical phenomena be the key to discovering laws that are still unknown to humanity? And can we discover them with the help of Artificial Intelligence?



ODSC gathers the attendees, presenters, and companies that are shaping the present and future of data science and AI. ODSC hosts one of the largest gatherings of professional data scientists with major conferences in USA, Europe, and Asia.