AI Solved The Three-Body Problem 100-Million Times Faster Than Anyone Before

avatar

The three-body problem seems quite simple. But really, it is not. The gravitational effects of three bodies can create an incredibly chaotic system.

geometry1848713_1920.jpg

Image by Gerd Altmann from Pixabay

It was Sir Isaac Newton who first formulated the Three-Body Problem back in the 17th century. And in essence, it is very simple. It is about how three space objects affect each other gravitationally. But soon it showed that the interactions are extremely complex. Three bodies create a chaotic system that is highly complex and highly sensitive to the initial settings of each of the three objects.

Currently, we use advanced software to solve the Three-Body Problem. But these programs can take weeks or even months to solve it. The most advanced of the programs is Brutus. And as its name suggests it uses brute force to solve the problem. But, we no longer need to use just brute force. Now we have AIs.

Chris Foley from the British University of Cambridge and his colleagues trained an AI that can solve the Three-Body Problem up to 100 million times faster than the current best programs – like Brutus. Such dramatic progress is awesome for astronomers. Now they can just ask the AI to help them with how the Universe is working.

To have a really useful AI you need to train it. In this case, the scientists used the current champion – Brutus – to generate 9900 simplified Three-Body Problem scenarios. They used these scenarios to train the AI than put it to work on another 5000 Three-Body Problems. The results were that the AI reached very similar results to Brutus. But where Brutus needs about 2 minutes the AI can solve the problem in a fraction of a second.

Traditional programs like Brutus are slow because they use brute force. The calculate each tiny step of the objects. On the other hand, AI observes the movement and derives how the system will look in future similar cases.

Using the AI for more complex Three-Body Problems will be much harder though. It will not be because the AI is too stupid to solve them though but preparing training scenarios on which it can learn will be insanely hard. But let's wish the scientists luck.

Sources:


  • If you like the content I’m producing about science maybe you will like the content I produce about gaming as well! Be sure to check out my other posts!


0
0
0.000
4 comments
avatar

Hello,

Your post has been manually curated by a @stem.steem curator.

FA8866FD-F2C3-43B3-A5A5-E0324BA4BB47.jpeg
Supporting Steemians on STEMGeeks

We are dedicated to supporting great content, like yours on the STEMGeeks tribe.

Please join us on discord.

0
0
0.000