An artificial intelligence tool has examined physical systems and, unsurprisingly, has found new ways to describe what it finds.
How do we understand the universe? There is no manual. There is no prescription.
At its most basic, physics helps us understand the relationship between “observable” variables—those are things we can measure. Speed, energy, mass, position, angle, temperature, charge. Some variables, like acceleration, can be reduced to more basic variables. These are the variables in physics that shape our understanding of the world.
These variables are bound together by equations.
Einstein’s most famous equation, E = Mike2summarizes the relationship between the variable energies (Second) and quality (Meter), use continuous: the speed of light (C). In fact, all of Einstein’s very complex special theory of relativity can be reduced to a relationship between three variables: energy, mass, and velocity.
There is nothing sacred about our choice of variables. The variables and mathematics we choose have stood the test of time because they make sense for a given theory or physical system.
But what if we were to find other physical variables to solve the same problem? It doesn’t change the problem…or the solution. But it may give us new insights into the inner workings of the universe and accelerate scientific discoveries.
Read more: Machine learning identifies origin of most famous Martian meteorite that landed on Earth
Now, artificial intelligence (AI) tools developed at Columbia University in New York have done just that.The experimental results were published in natural computational science.
Robotic experts at Columbia Engineering developed an AI program to review raw video data and search for the smallest set of fundamental variables that fully describe the observed physical dynamics of the system, the swing of the pendulum.
To test their artificial intelligence, the team first showed a tool video of a phenomenon they already knew the answer to.
The double pendulum can be described by four “state variables” – the angle and angular velocity of each of the two arms. After staring at the video for a few hours, the AI gave the answer for the number of system variables: 4.7.
“We think the answer is close enough,” said senior author Hod Lipson, director of the Creative Machines Laboratory in Columbia’s Department of Mechanical Engineering. “Especially since all the AI has access to raw video footage without any knowledge of physics or geometry. But we wanted to know what the variables actually were, not just their numbers.”
So the next challenge was to try to visualize the variables identified by the AI. This is not easy because the program does not describe the variables in human-intuitive language. However, the researchers did associate two of these variables with the angle of each swing arm.
“We tried to relate other variables to anything we could think of: angular and linear velocities, kinetic and potential energies, and various combinations of known quantities,” explained lead author Boyuan Chen, now an assistant professor at Duke University. “But nothing seems to be a perfect match.”
Boyuan Chen, now an assistant professor at Duke University, said the team tried to relate other variables to anything they could think of: angular and linear velocities, kinetic and potential energies, and various combinations of known quantities. “But nothing seems to be a perfect match.”
Convinced that the AI’s good predictions of system dynamics meant it had found a valid set of four variables, the team was confused about what the other variables might be.
“We don’t yet understand the mathematical language it speaks,” Chen said.
Still, AI has returned good computational results on other physical systems with known solutions.
Confused already, what does the team lose if they show the AI something for which there is no known answer?
The team showed the AI a video of a “weird, wavy, inflatable arm swinging man,” or family man Fame, tossing in front of used car yards. A few hours of analysis returned eight variables. The Lava Lamp video also gives eight variables. A looping video of a fireplace yields 24 variables.
The researchers wanted to know if the variable set was different each time the program was restarted, or if the same unique set of variables was found for each system.
“I’ve been wondering, if we met an intelligent alien race, would they discover the same laws of physics as we do, or would they describe the universe differently?” asked Lipson. “Perhaps some phenomena seem very complex because we’re trying to understand them using the wrong set of variables.?”
In the experiment, the number of variables remained the same, but each time the AI restarted, the specific variables changed. This suggests that there are alternative ways of describing the system – and by expanding the universe – that our choices may not be perfect.
This AI tool could help scientists grasp complex phenomena that current theories cannot understand. “While we used video data in this work, any type of array data source could be used — such as radar arrays or DNA arrays,” explains co-author Kuang Huang.
Lipson argues that scientists can misunderstand or fail to understand many phenomena simply because they don’t have a good set of variables. “For thousands of years people have known how fast or slow objects move, but it was only when the concepts of velocity and acceleration were formally quantified that Newton discovered his famous laws of motion F = mother. “
Likewise, temperature and pressure variables need to be described before the laws of thermodynamics can be formalized. This is true of any scientific theory – to develop a theory, you first need variables. “Just because we don’t have variables, what other laws are we missing?” asked co-author Qiang Du, a math professor.