AI Just Found Alternative Physics Independently

Pick up any physics textbook and you’ll find formula after formula that explains how things sway, fly, turn, and stop. The formula describes actions that we can observe, but behind each can be a series of factors that are not immediately apparent.

Now, a new AI program developed by researchers at Columbia University appears to have found its own alternative physics.

After being shown videos of physical phenomena on Earth, AI did not rediscover the current variables we used; instead, it actually comes up with a new variable to explain what it sees.

To be clear, this does not mean that our current physics is flawed or that there is a more suitable model to explain the world around us. (Einstein’s laws have proven to be very powerful.) But they can only exist because they are built on a pre-existing ‘language’ of theories and principles established by centuries of tradition.

Given an alternate timeline where other minds tackle the same problem from a slightly different perspective, would we still frame the mechanisms that explain our Universe in the same way?

Even with newer technology imaging black holes and detecting strange and distant worlds, these laws have held up time and time again (side note: quantum mechanics is another story, but let’s stick to the visible world here).

This new AI only sees videos of a handful of physical phenomena, so it’s not at all suited to coming up with new physics to explain the Universe or trying to do the best of Einstein. This is not the goal here.

“I’ve always wondered, if we ever met a race of intelligent aliens, would they discover the same laws of physics as we have, or could they describe the universe in a different way?” says robotic expert Hod Lipson of Creative Machines Lab in Columbia.

“In the experiment, the number of variables was the same every time the AI ​​was restarted, but the specific variables were different each time. So yes, there is an alternative way of describing the Universe and it’s very possible that our choices aren’t perfect.”

Beyond that, the team wanted to find out if AI could actually discover new variables – and therefore help us explain complex new phenomena emerging in our current flood of data that we currently have no theoretical understanding to follow.

For example, new data emerging from giant experiments like the Large Hadron Collider hint at new physics.

“What other law are we missing just because we don’t have a variable?” says mathematician Qiang Du of Columbia University.

So how did AI discover new physics? To begin with, the team entered the system’s raw video footage of the phenomena they already understood and asked the program a simple question: What are the minimum fundamental variables needed to describe what happened?

The first video shows a swinging double pendulum that is known to have four state variables: the angle and angular velocity of each of the two pendulums.

The AI ​​pondered over the footage and questions for several hours and then provided an answer: This phenomenon would require 4.7 variables to explain it, he said.

That’s pretty close to the four we know of… but it still doesn’t explain what the AI ​​thinks about that variable.

So the team then tried to match the known variables to the variables that the AI ​​had chosen. Two of them loosely fit the angle of the arm, but the other two variables remain a mystery. However, AI was able to make accurate predictions about what the system would do next, so the team guessed that the AI ​​must be doing something they couldn’t understand.

“We tried to correlate other variables with anything and everything we could think of: angular and linear velocities, kinetic and potential energies, and various combinations of known quantities,” said software researcher Boyuan Chen, now an assistant professor at Duke University, who led the research. profession.

“But nothing seems to fit perfectly … we don’t yet understand the mathematical language it uses.”

The team then proceeded to show another AI video. The first featured an ‘air dancer’ of wavy arms blowing in the wind (AI says this has eight variables). The lava lamp recording also yielded eight variables. The fire video clip returns with 24 variables.

Each time, the variable is unique.

“Without prior knowledge of the underlying physics, our algorithm discovers the intrinsic dimensions of the observed dynamics and identifies a candidate set of state variables,” the researchers wrote in their paper.

This suggests that in the future, AI has the potential to help us identify variables that underpin new concepts that we are not currently aware of. Pay attention to this space.

This research has been published in Natural Computational Science.

#Alternative #Physics #Independently