World’s First AI Universe Simulator Works So Well It’s Scary - Physics-Astronomy.org

World’s First AI Universe Simulator Works So Well It’s Scary





As traveling back in time billions of years is not yet a possibility, scientists have used computer simulations to try and digitally reverse-engineer the origin and evolution of our universe using what we do know about it.
Till now, these simulations fell into one of two categories: slow and more accurate, or fast and less accurate. But now, an international team of researchers has pioneered an AI simulator that can quickly generate highly accurate, three-dimensional simulations of the universe, baffling its designers by its ability to understand things about the cosmos that it wasn’t trained on.
Researcher Shirley Ho said in a press release, “It’s like teaching image recognition software with lots of pictures of cats and dogs, but then it’s able to recognize elephants. Nobody knows how it does this, and it’s a great mystery to be solved.”
The scientists describe how they created this universe simulator christened the Deep Density Displacement Model (D3M), in a study that was recently published in the journal Proceedings of the National Academy of Sciences.




The aim was to teach D3M how to model the way gravity shapes the universe. To achieve this goal, they fed the system 8,000 different gravity-focused simulations created by a highly accurate existing universe simulator. The system needed 300 computation hours to generate just one of its simulations, but after training on the data, D3M was able to produce an impressively original simulation of a cube universe 600 million light-years across in just 30 milliseconds. These simulations were more accurate than those of the existing “fast” systems, which still needed a couple of minutes to create a simulation.
Amazingly, speed isn’t the most remarkable thing about D3M. Its capability to accurately simulate what the universe would look like even when the researchers changed parameters that weren’t included in its training data emerged as its most radical feature. For instance, the team didn’t train the system on data on universe with varying amounts of dark matter, yet the AI inexplicably changed those values based on inferences from the ones it was trained on and could accurately simulate that universe’s evolution.
In addition to helping astrophysicists and researchers like Ho better understand the universe’s evolution, this unpredictable behavior of D3M has the potential to help computer scientists better understand AI.
Ho said in the press release, “We can be an interesting playground for a machine learner to use to see why this model extrapolates so well, why it extrapolates to elephants instead of just recognizing cats and dogs. It’s a two-way street between science and deep learning.”

1 comment:

  1. I can’t believe this. A great testimony that i must share to all HERPES patient in the world i never believed that their could be any complete cure for HERPES or any cure for HERPES, i saw people’s testimony on sites of how DR Afrid prepare herbal cure and brought them back to life again. i had to try it too and you can, t believe that in just few weeks i started using it all my pains stop gradually and i had to leave without the HERPES DRUGS the doctor gave to me. Right now i can tell you that few months now i have not had any pain, delay in treatment leads to death. Here is his email: drafridherbalhome@gail.com his WHATSAPP number is +2349057260738.

    ReplyDelete

Powered by Blogger.