|The Leading Source for Global News and Information Covering the Ecosystem of High Productivity Computing / November 15, 2005|
Geordie Rose can't wait for the future -- so, apparently, he's
not going to. Rose is president and CEO of D-Wave Systems Inc., a
company with the ambitious goal of building a quantum computing system
for the commercial market. In this exclusive interview, HPCwire editor
Michael Feldman spoke with Geordie Rose to discuss his
company and what he intends to offer the HPC community.
HPCwire: Could you briefly describe how a quantum computer operates and how it is fundamentally different from that of a conventional computer?
GEORDIE ROSE: One of the most profound realizations of the previous century is that computation and physics are deeply connected -- all of the stuff that makes up our world is constantly "computing", that is, evolving in time, and that when we build computers we are just tapping into that ongoing computation.
When a computer is built out of devices that behave according to the rules of classical physics, like PowerPCs, Pentiums, or balls rolling down hills, some of the computing resources that nature allows are not used. Classical physics can be thought of as a highly restricted version of quantum physics. This restriction removes computational capabilities that are allowed by quantum physics but not classical physics.
When computers are built out of components that can harness the resources made available by quantum mechanics, they gain access to computational resources that are not available to conventional computers. One analogy is to think of computation as painting. In this analogy, completing a computation is like completing a painting. A classical painter only gets to use black and white paints, but a quantum painter gets to use red, yellow, black and white. If the job is to paint a picture of the sun, the classical painter can never get it right, no matter how skilled, because he doesn't have access to the right resources; that is, red and yellow paint don't exist for him, while the quantum painter finds it pretty easy.
HPCwire: Could you present some real-world examples of problems that can be addressed by quantum computers but are impossible to solve with conventional computers?
ROSE: Yes. But I should point out that this capability is not required to make quantum computers worth building. It is likely in the short term that they will be cheaper, easier to use, and just a little better on performance than enterprise-class supercomputers and this will justify their build-out from a commercial perspective.
The main application where "exponential" improvements occur using quantum computers over classical ones is something called quantum simulation, which is the "in silico" prediction of the properties and/or behavior of an atomic or molecular scale product or process. Even very small quantum computers can outperform the most ambitious supercomputers ever conceived in these tasks because of the inherent advantage of quantum computers for this task.
For many industries, predictive modeling of atomic and molecular scale products and processes would be transformative, including the chemical, biotech, energy and pharmaceutical industries.
HPCwire: There seems to be a general consensus that quantum computing is still in its pioneering stage. What events compelled you to attempt commercialization at this time?
ROSE: It is undoubtedly true that quantum computing was in its infancy when we incorporated D-Wave in 1999. However there were some very good arguments for believing, back in 1999, that the timing was right to structure an organization whose objective was to gain and hold a leadership position in the development and commercialization of quantum computing.
Two of the principle drivers for starting the company when we did were the following: First, while there were literally hundreds of ongoing projects in universities and corporate labs studying aspects of quantum computing, there were only a handful of real efforts trying to build quantum computers. This meant that our effort stood a real chance to become the leader in this field, which we successfully accomplished. Second, even back in 1999, nearly every scientist who had thought carefully about quantum computing technology believed that eventually they would be built -- the question was when, not if. The corollary was that if we started to attack the problem early, using best practices from the semiconductor industry and careful selection of top engineers, management, investors and scientists, we could set up a high-throughput industrial-strength infrastructure to accelerate this inevitability.
HPCwire: Could you describe the overall design of D-Wave's quantum computer?
ROSE: The system we are currently deploying, which we call Trinity, is a capability-class supercomputer specifically designed to provide extremely rapid and accurate approximate answers to arbitrarily large NP-complete problems. This class of computational problem is common in naturally discrete problems with complicated networks of interactions, such as those found in some life sciences, cognition, logistics, electronic design automation and quantitative finance applications.
Trinity has a front-end software interface, implemented in a combination of Java and C, that allows a user to easily state any NP-complete problem of interest.
After such a problem has been stated the problem is compiled down to the machine language of the processors at the heart of the machine. These processors then provide an answer, which is shuttled back to the front end and provided to the user. This capability can of course be called remotely and/or as a subroutine of some other piece of software.
The processors themselves are made out of lithographically defined patterns of niobium (a metal at room temperature, a superconductor when cooled) on a standard inert substrate, such as quartz. They are analog, software programmable circuits designed to harness the most robust quantum resources to accelerate computation. In the "computation is like painting" analogy, these processors can paint with yellow, but maybe not red. So you can paint different pictures than the classical black-and-white painter, but you might not be able to paint in red, even though this is an allowed quantum resource.
This is a subtle but important point that many discussions about quantum computing ignore. It is not true that there is a sharp dividing line between what is a classical and what is a quantum computer. Machines can be built that are somewhere in the middle that avoid most of the difficult engineering problems of building entirely quantum computers, but still harness quantum mechanics to massively accelerate computation of high-value, real-world problems.
HPCwire: When do you expect to have a working prototype?
ROSE: The first Trinity prototype will be running our clients' problems by March 2006.
HPCwire: How do you think the arrival of quantum computers will change the landscape of the high-performance computing industry?
ROSE: This is difficult to predict. I suspect that in the short term Trinity will be viewed as a competitor for the current enterprise- and capability-class supercomputer market for the restricted segment of the market Trinity is targeted at -- which includes NP-complete problems, monte carlo simulations and quantum simulation problems.
However Trinity differs significantly from competing supercomputer technologies in several important ways. They are cheap to build; the performance of the machines is expected to scale far beyond current limitations; Trinity generates relatively little heat during its operation because the bulk of the computation is done in superconducting electronics; Trinity takes up very little physical space; the machine is very easy to program. All of these differences will likely lead to further segmentation of the HPC market classifications, drawing market share from the enterprise- and capability-class segments and re-assigning it to the new class of machine that Trinity is. In addition of course we hope to expand the market size based on the new capabilities of Trinity systems over other capability-class supercomputers.
While at SC05, Geordie Rose will be conducting a Masterworks session on quantum computing. It will take place on Wednesday, November 16, from 4:15 pm to 5:00 pm. For more information visit http://sc05.supercomputing.org/schedule/event_detail.php?evid=5311