New Study: Algorithms based on deep neural networks can be applied to quantum physics

Print Friendly, PDF & Email

Intel Senior Vice President and Mobileye CEO Prof. Amnon Shashua

A computer science research group from the Hebrew University of Jerusalem has mathematically proven that artificial intelligence (AI) can help us understand currently unreachable quantum physics phenomena. The results have been published in Physical Review Letters. Intel Senior Vice President and Mobileye CEO Prof. Amnon Shashua shared the group’s findings during a keynote Wednesday at the Science of Deep Learning Conference hosted by The National Academy of Science in Washington, D.C.

Our research proves that the AI algorithms can represent highly complex quantum systems significantly more efficiently than existing approaches,” said Prof. Amnon Shashua, Intel senior vice president and Mobileye president and CEO.

Despite the surge of AI across nearly every industry, it has not been widely applied to the world of quantum physics. Doctoral students Yoav Levine, Or Sharir and Nadav Cohen, led by Shashua, aim to change that by proving how recent developments in machine learning can help us study some of the previously unreachable areas of quantum physics. Using the latest advancement in deep neural networks to conduct proper simulations faster and more thoroughly, these researchers argue, will provide new insight into the smallest of particles and how they interact.

A deeper view into this area of physics has the potential to unlock the next revolutions in computing, energy and transportation. For example, quantum computing, an area Intel has invested more than $50 million in to research, could advance further with a more efficient approach to testing highly complex quantum systems.

The team showed that algorithms based on deep neural networks – algorithms that have revolutionized AI – can be applied to the world of quantum physics. These algorithms, which have already endowed computers with facial- and voice-recognition capabilities, will now be able to refine our understanding of the quantum behavior of nature.

Understanding phenomena in systems of many interacting quantum particles (particles of miniscule size, such as electrons) is one of the hottest topics in contemporary physics research. However, even the most experienced researchers are unable to gain more than a mere glimpse of these complex phenomena. Due to the enormous amounts of particles (over a billion billions in one gram) and the myriad interactions between them, it is extremely difficult to conduct a simulation that will allow a comprehensive understanding of these systems. Even the strongest available computer programs have difficulties with this challenge, which seemed uncrackable. Until now.

Many-particle quantum physics is one of the most popular and intriguing subfields in current physics research. It studies how particles in nature “come together” and bring forth surprising properties, such as electrical conductivity and magnetism, among others. As has happened in the technological revolutions of the 20th century, a deeper understanding in this domain can greatly affect various aspects of modern life, as it bears potential to enable the next revolutions in computing, energy, transportation – and the limits can only be imagined. The connection of AI to this field promises fascinating developments in upcoming years.

Sign up for our insideHPC Newsletter