Short Bytes: “In future, you might hear about intelligent machines that’ll learn new things on their own and help create a full-fledged quantum computer.” This prediction is inspired by the recent work of researchers that involves the use of neural networks to understand the quantum many-body problem. The AI used in this research was similar to the one that conquered the ancient game of Go.Do you remember how Google’s Alpha Go artificial intelligence neural network mastered the ancient game of Go and beat world champions? It looks like the same technology can be utilized to solve some other tricky modern problems. And, what could be trickier that understanding the quantum physics?
In the past, traditional methods to understand the behavior of quantum interacting systems have worked well, but there are still many unsolved problems. To solve them, Giuseppe Carleo of ETH Zurich, Switzerland, used machine learning to form a variational approach to the quantum many-body problem.
Before digging deeper, let me tell you a little about the many-body problem. It deals with the difficulty of analyzing “multiple nontrivial relationships encoded in the exponential complexity of the many-body wave function.” In simpler language, it’s the study of interactions between many quantum particles.
If we take a look at our current computing power, modeling a wave function will need lot more powerful supercomputers. But, according to Carleo, the neural networks are pretty good at generalizing. Hence, they need only limited information to infer something. So, fiddling with this idea, Carleo and Matthias Troyer created a simple neural network to reconstruct such multi-body wave function.
“I like saying that we have a machine dreaming of Schrödinger’s cat.”
By testing some sample problems, they were able to know that it performed better than other available tools. They also calculated the lowest energy or ground states.
“It’s like having a machine learning how to crack quantum mechanics, all by itself,” Carleo says, according to New Scientist. “I like saying that we have a machine dreaming of Schrödinger’s cat.”
This work has sparked a new idea of using neural networks to create an efficient representation of quantum systems. With the increasing advancements in machine learning, we can surely get more insights into intricate problems. And, one day, who knows, an AI-powered supercomputer might just create a quantum computer on its own!
Did you find this story on the use of neural networks to solve quantum problems interesting? Don’t forget to share your feedback.