Simulating the universe is difficult.

There are, after all, potentially infinite variables to consider.

It could very well change the way we perceive the universe and understand the laws of physics.

Carnegie Mellon researchers trained AI to simulate our universe on a GPU

40% off TNW Conference!

This is because physics is still an open problem.

And that means we have to make crap up as we go along.

In this way its a massive feat of trial-and-error.

The problem

Running a supercomputer is expensive.

Just renting one costs upwards of thousands of dollars an hour.

Supercomputers arent especially good solutions for processing problems that intentionally require trial-and-error.

This is a work-stopper when it comes to attempting to simulate the entire cosmos.

The solution, of course, is AI.

And this makes things faster.

With the new approach, the researchers need only 36 minutes.

The results were even more dramatic when more particles were added to the simulation.

Using current methods, a simulation of this size and resolution would take a dedicated supercomputer months to complete.

This doesnt mean the AI knows what the universe beyond our reach looks like.

What we dont know is exactly how it got this way.

you could read the whole studyhere.

Also tagged with