In mid-November, NTT Research issued a press release announcing that the PHI Lab, which I direct, had reached five-year joint research agreements with eight other research organizations: six universities, one U.S. government agency and one quantum computing software company. In this brief article, I would like to describe some of the background behind this initiative and share some more insight on what we are hoping to achieve.
The focus of the PHI Lab is to discover a new computational interface between quantum physics, neuroscience and optical technology. This conceptual model, called a Coherent ising machines (CIM), is not exclusively our own. Nor could it be, because NTT does not cover the vast area of science that CIM entails. For the past several decades, we have been working primarily in one of these areas, the optical domain. One of NTT’s early contributions relevant to this initiative was our proposal, about 40 years ago, of coherent optical communications. Our more recent contributions include our work in Optical Parametric Oscillators (OPOs) as a phase sensitive amplifier and Coherent Ising Machines (CIMs).
To this joint effort, we bring our individual strengths, as do our eight collaborators. Their primary goals in this research initiative include the following;
• California Institute of Technology (Caltech): to develop scalable architecture for efficient quantum simulation of many-body spin systems using OPO networks
• Cornell University: to develop a k-SAT solver based on error detection and error correction feedback
• University of Michigan: to perform theoretical studies of topological states with anyon statistics in nonlinear optics and synthetic topological matter
• Massachusetts Institute of Technology (MIT): to develop the photonic accelerators for deep learning and the superconducting CIMs for fundamental study
• NASA Ames Research Center: to perform benchmark studies of CIMs vs. modern heuristics on various optimization problems
• Stanford University: to develop novel optical and superconducting devices for studying the quantum-to-classical crossover physics and critical phenomena in the neural network
• Swinburne University: to develop and implement the quantum phase space methods for CIMs
• 1QBit: to perform research in design and analysis of a stack of algorithms that bridge commercially viable applications to the forms of computation natively done by CIMs, with a multitude of applications in operations research and artificial intelligence
As you can see, this is not a single-purpose mission. We are covering a broad expanse, ranging from quantum physics to algorithm development, and even drawing inspiration from how the brain works. Our work with OPOs, which can act as both quantum and classical processors, places us outside the mainstream of quantum computing work today. But as our list of collaborators indicates, we are not alone. And as certain weaknesses in conventional quantum computing become more widely recognized, we expect interest in the CIM approach will grow.
In any case, and however much attention this effort may generate, our goals are more fundamental. We aim actually to conduct the kind of basic, cutting-edge research, which over a number of years, will uncover the very nature of computation.
