- American Physical Society Sites
- Meetings & Events
- Policy & Advocacy
- Careers In Physics
- About APS
- Become a Member
By Leah Poffenberger
After its cancelation last year due to the coronavirus pandemic, the March Meeting came back strong in 2021, with a new online format drawing in a record number of March attendees—over 13,000 registrants. While many aspects of the meeting might have differed from a traditional March Meeting, the Kavli Foundation Special Symposium returned as a highlight of the meeting, featuring distinguished researchers in the fields of quantum computing and machine learning.
The five speakers on hand to share their expertise in these fields were: Patrick Francis Riley (Google Accelerated Science), Roger Melko (University of Waterloo, Perimeter Institute), Michelle Girvan (University of Maryland, College Park), Eun-Ah Kim (Cornell University), and John P. Preskill (Caltech).
After a brief introduction from APS CEO Jonathan Bagger, who highlighted the “potential of [machine learning and quantum computing] to transform physics,” Riley kicked off the Kavli symposium. In his talk “Vignettes of Machine Learning in the Natural Sciences,” Riley detailed on-going projects at Google Accelerated Science to harness machine learning technology. First, he described the use of machine learning as a tool for small molecule drug discovery, allowing researchers to search for potential therapeutic molecules on a previously impossible scale. He also presented another project that seeks to integrate computational physics and machine learning by using differentiable programming to tackle physics problems like solving Schrödinger’s equation.
Next, Melko gave a talk on “Machine Learning and the Complexity of Quantum Simulation,” where he discussed the use of generative models from machine learning to enhance quantum simulations. Melko also discussed two forms of quantum simulation—Hamiltonian-driven and data-driven—along with the difficulties that surround each method. With data-driven generative modeling, Melko says we are “entering a new era” of simulation, which could lead to discoveries in condensed matter and quantum information. Generative modeling has the potential to increase our understanding of the complexity of quantum simulation.
Girvan followed up, with a talk on more uses for machine learning technology titled “Opening the black box: Improving knowledge-free machine learning with knowledge-based models.” In her work, Girvan combines artificial intelligence and machine learning approaches with mathematical modeling, with the goal of predicting complex, chaotic systems such as the weather, and stochastic systems with randomness, such as the stock market. Using a subset of machine learning called reservoir computing and incorporating “knowledge”—such as standard rules of physics—into a hybrid scheme, Girvan showed good predictions of complex weather systems.
Kim discussed ways to harness machine learning for the understanding of quantum emergence. In order to develop theoretical insight from complex experimental data, or to make predictions based on this theoretical insight, Kim has worked with several groups using machine learning tools in various ways, such as hypothesis testing to better understand quantum matter. Approaches like neural networks can solve problems related to a huge volume of data available in the study of quantum emergence. According to Kim, different types of machine learning can provide new insights into complex discoveries or accelerate discovery by processing huge amounts of data in a short period of time.
To close the Kavli Symposium, Preskill provided an overview of the field of quantum information science in his talk “Quantum Computing: current status and future prospects.” He discussed the potential strengths of quantum computing, the difficulties involved with developing quantum systems, and current quantum devices that are producing interesting results. According to Preskill, the field is currently in the noisy intermediate-scale quantum (NISQ) era, which is suited for scientific exploration, but experiences limited computational power due to noise limits. Quantum researchers are now setting their sights on moving from NISQ to fault-tolerant quantum computers, which will exponentially scale up the number of physical qubits but will also have direct applications to physics, chemistry, materials science, and more.
©1995 - 2021, AMERICAN PHYSICAL SOCIETY
APS encourages the redistribution of the materials included in this newspaper provided that attribution to the source is noted and the materials are not truncated or changed.
Editor: David Voss
Staff Science Writer: Leah Poffenberger
Contributing Correspondents: Sophia Chen, Alaina G. Levine