Published on : Jun 28, 2019
Astrophysicists have deployed artificial intelligence techniques for the first time to create highly complex 3D simulations of the universe. The corresponding results are so robust, accurate, and fast that it has even surprised the developers. The scientists are unsure of how it all came together.
Researchers have run these simulations, which give results in few milliseconds. On the other hand, some other simulations may take up to couple of minutes. Interestingly, the accuracy of the results is quite high given the time taken to complete the simulations.
The project is called D3M – acronym for Deep Density Displacement Model. The accuracy and the speed of the project was not the cause of the biggest surprise for the creators. They were really taken aback by the accuracy showed by the model to simulate how the universe would look if a few parameters were changed. The model was not offered any sort of training data for such variable parameters.
Such computer simulations are now an integral part of theoretical astrophysics. Researchers want to explore how the universe may evolve under different scenarios. For instance, how dark energy may pull the cosmos apart over time. Such complex research need hundreds of simulations to clearly understand the matter. Having a computer model that yields fast and highly accurate results will only help scientists in unravelling the mysteries of the universe.
The D3M can simulate how gravity influences the cosmos. The scientists have chosen to focus only on gravity as it is the most important force that has influenced the large-scale evolution of the universe, by a mile.
The study is led by a Flatiron Institute research analyst, Siyu He. The research is co-authored by Shirley Ho who is a group leader at the Flatiron Institute Center in New York. She also works as an adjunct professor at the Carnegie Mellon University. He and his team published the D3M and its findings in the Proceedings of the National Academy of Sciences.