Artificial Intelligence Uncovers Alternative Physics by Observation, Study Claims

Columbia Engineering researchers put out a principle to figure out how many state variables and what kind of variables may be present in an observable physics system.

A machine learning (AI) system was taught by Dr. Boyuan Chen and his co-authors to count the number of variables required to describe physical systems and forecast advancements. A report on the findings may be found in Nature Computational Science.

"I always wondered, if we ever met an intelligent alien race, would they have discovered the same physics laws as we have, or might they describe the universe in a different way?" Study senior author Professor Hod Lipson in a statement (via Eurekalert). "Perhaps some phenomena seem enigmatically complex because we are trying to understand them using the wrong set of variables."

How Cosmos Would Look Like With Another Parameter

Scientists occasionally catch a brief peek of what the cosmos may look like if the universe had started with a different set of parameters.

The conventional triangle variables of length and angle were changed by mathematician Norman Wildberger to squares of the length and sine of the angle, which he refers to as quadrants and spread, to produce what he terms "rational trigonometry."

When these variables are used, specific problems become significantly simpler. Still, for someone who has studied Euclidean geometry, this initially feels like speaking a different language.

Some civilizations, most notably the native American Hopi, are said to see elements like time in a way that is fundamentally different from the majority of the rest of the world.

Given that raising a child in such a way is "against physics," IFL Science said the researchers resorted to AI, beginning with a video of elastic double pendulums.

The angle and angular velocity of each arm are most likely the four variables that a physicist perceives while studying a double pendulum system. We can simply quantify the four, and they make sense to other people.

Physics Students Use AI to See How Variables Would Change

Undergraduate physics students are taught to represent the system in terms of each arm's kinetic and potential energy.

The scientists queried a neural network about how many state variables it perceived after showing it a video of a double pendulum, SciTechDaily said.

Even though the answer was four, neither the machine nor the people could agree on what these variables were since they didn't have a common language. Two seem comparable to how we calculate arm angles, while the other two are still a mystery.

The authors then demonstrated much more intricate dynamical systems to the computer, including flames in a fireplace, a lava lamp, and an "air dancer" outside a nearby car dealership. It claimed that eight, eight, and 24 state variables were required to describe these systems. Still, nobody is yet sure of what these variables are.

The dynamics of physical systems have been modeled by earlier machine learning tools. Still, they relied on measurements of relevant state variables, or quantitative variables that accurately capture the system as it changes. After being instructed in this manner, the machines were unlikely to think of additional variables independently.

Now, it appears that AI systems can recognize novel variables; all we need is a translator to determine what they are.

Check out more news and information on Physics in Science Times.