Recently, a team of researchers pioneered the world’s first AI (Artificial Intelligence) simulator. It’s quick, and its creators are confused by it’s ability to understand things about the Universe that it should not.
Scientists used computer simulations to try to reverse the origin and evolution of our universe digitally for decades. The best traditional methods using modern technology take minutes and produce Good results. The world’s first AI universe simulator, on the other hand, produces results with much greater precision in just milliseconds.
According to the team’s paper:
Here, we build a deep neural network to predict the formation of the structure of the Universe. It overcomes the traditional analytical quick approach and accurately extrapolates far beyond its training data.
This is a sophisticated way of saying that it does not just do what its developers have created it to do – simulate the evolution of the universe under different gravitational conditions – as it produces accurate results for variables over which it has not been trained. For example, a particular parameter that the reported simulation surprised the scientists was the amount of dark matter in the universe.
The team did not train the system, called Deep Density Displacement Model (D3M), on data with varying amounts of dark matter, but AI inexplicably (and, according to the research, precisely) modified those values.
As Shirley Ho, a co-author on the role of team researcher said:
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 simulator itself, aside from further demonstrating the Mercurial and unpredictable nature of black box AI and deep learning, has the potential to help astrophysicists and researchers fill in some of the blanks in our comsos’ backstory.
Our universe is a strange and almost unknown place. Humanity is just beginning to define our visions beyond the observable space to determine what is out there and how it all turned out the way it is. Artificial intelligence can help us understand exactly how billions and billions of variables that affect the evolution of our universe work in the emergence of stars, planets and even life itself.
The AI was developed by the lead author of the team’s work, Suyi He of the Flatiron Institute and Carnegie Mellon University, along with co-authors Yin Li of UC Berkeley and Kavli Institute, Yu Feng of UC Berkeley, Shirley Ho of Flatiron Institute and Carnegie Mellon Univesity, Wei Chen of the Flatiron Institute, Siamak Ravanbakhsh of the University of British Columbia in Vancouver, and Barnabas Póczos of Carnegie Mellon University.