Grant: Discovery Grants Program - Individual

Computer simulations of quantum matter enable researchers to discover and design new molecules, materials, and devices. A key challenge in our efforts to simulate quantum matter is limitations in numerically modelling the complex correlations that occur in quantum systems. This has motivated the development of experimental quantum devices with the ability to synthesize quantum matter with hundreds -of qubits in the lab. It is intriguing to inquire if such devices can be used to accelerate solutions to hard computational problems, such as the simulation of superconductivity, topology, or quantum phase transitions. Concurrent to the rapid development of these experimental quantum devices, a new generation of technology based on machine learning has shown unique abilities in a wide variety of computational tasks. Breakthroughs in functionality, such as those seen in the rise of large language models, raise a number of exciting scenarios surrounding the interplay between quantum and classical simulation strategies. Most obviously, machine learning is well--suited to leverage the large amounts of qubit measurement data produced by a quantum device.  In turn, machine learning models can be trained to guide the design of the quantum device, opening up the possibility of new research paradigms in the future. In this proposal, we aim to leverage our world--leading research program in the development of language models for quantum simulation, to make progress on a number of scientific questions related to the learning of quantum states of matter. Our main objective is to develop powerful language models to apply to the current generation of experimental quantum devices. We will explore whether machine learning models are able to learn quantum states, and determine when the design of quantum computers can be accelerated by using language models.  We will determine whether large language models display emergence of macroscopic quantum phenomena, like topological order or exotic quantum criticality, if scaled sufficiently large. Finally, we will develop meaningful connections between machine learning models and the foundations of physics through quantum field theory. This research program will train a diverse workforce of researchers, and provide a unique training environment that includes a traditional physics education, as well as industrial internships and entrepreneurial mentorship. Our research will provide significant benefits to Canada's nascent quantum computing industry, including the development of quantum computing expertise through collaboration with quantum computing companies. Given the extraordinary abilities demonstrated recently by machine learning in language processing tasks, our team will explore what large language models could achieve in physics, both in scaling quantum computing, and also in advancing the fundamental theoretical understanding of quantum matter, materials and devices.