Scientists build incredibly accurate universe simulation — but don’t know how they did it


Everyone’s had a moment or two in their lives where they pull off an incredible feat but can’t explain how they did it. Perhaps it was a near-superhuman catch in a softball game or acing a test without studying. In any case, it’s happened to all of us, and now a group of astrophysicists is having to deal with their own unexplained good fortune in a very public way.

The international team of scientists built a three-dimensional simulation of the universe that is both incredibly fast and remarkably accurate. The only problem? They have no idea why it works as well as it does.

The work, which is the subject of a new paper published in Proceedings of the National Academy of Sciences, represents an incredible leap in universe simulations, and could provide scientists with more power to understand the nature of, well, everything.

The system is capable of conducting its simulations much more rapidly than its predecessors, but that’s really only half of the story. The AI-powered model is also capable of simulating what the universe might be like under many different conditions, and the scientists who created it are having a hard time explaining how this is possible.

“It’s like teaching image recognition software with lots of pictures of cats and dogs, but then it’s able to recognize elephants,” Shirley Ho, co-author of the study, said in a statement. “Nobody knows how it does this, and it’s a great mystery to be solved.”

The system, called the Deep Density Displacement Model, could quickly become a topic of study for scientists hoping to better understand the intricacies of deep learning. For astrophysicists, it could become a useful tool in future research, providing more accurate results than the alternatives.