“We tried to correlate the other variables with everything we could think of: angular and linear velocities, kinetic and potential energy, and various combinations of known quantities,” but nothing seemed to fit perfectly, the scientists say.
Researchers at Columbia University in New York have developed a new artificial intelligence (AI) algorithm that has apparently discovered its own alternative physics. The scientists presented the AI with videos of a variety of physical systems, with the aim of determining how many hidden state variables these systems are likely to have, and what they might be. However, they were puzzled by the results, as the AI identified new candidate sets of totally unsuspected state variables, reported scientists this Monday on the university portal.
After showing him videos of physical phenomena on Earth, ranging from elastic double pendulums to flames in a fireplace, the AI did not confirm the variables used in physics to describe these systems, which are already understood from existing theories, but new variables came up to explain them.
“We try to correlate the other variables with everything that is happened: angular and linear velocities, kinetic and potential energy, and various combinations of known quantities […] but nothing seemed to fit perfectly,” said Boyuan Chen, assistant professor, and software expert at Duke University, who led or research.
Every time the AI was restarted, the number of fundamental variables needed to explain the system was the same, but they were different each time. The researchers concluded that there are alternative ways of describing the universe and it is quite possible that such approaches ummed by science are not the correct ones.
The experts stressed that the designed algorithm discovered the intimate dimension of the observed dynamics and identified candidate sets of state variables, without any prior knowledge of the underlying physics. They also suggested that, in the future, AI could help identify variables that support new concepts of which there is currently no knowledge.
“What other laws are we missing simply because don’t we have the variables?” asked mathematician Qiang Du of Columbia University and co-author of the study. The findings were presented in the journal Nature Computational Science.