When the theorists can’t agree, experimentalists forge ahead. Because the whole point of adiabatic quantum computation is to go slow rather than fast, adiabatic quantum computers are in principle significantly easier to build than general-purpose code-breaking quantum computers. Realizing this key point, in 2002 my graduate student Bill Kaminsky and I created a design for an adiabatic quantum computer based on superconducting technology. Last year, D-Wave Systems, a quantum-computing startup in Burnaby, British Columbia, announced that it had constructed an adiabatic quantum computer based on our design. At that point, things got interesting.
D-Wave was founded a little less than a decade ago, with the express purpose of building a commercial quantum computer. After toying with the idea of building a quantum computer to factor large numbers, its researchers sensibly settled on the more straightforward and still potentially profitable tasks of quantum simulation and adiabatic quantum computing. In February 2007, at Silicon Valley’s Computer History Museum, the company demonstrated a 16-qubit device that it claimed could solve reasonably complex optimization problems. It could even do Sudoku puzzles!
D-Wave has raised about $60 million in funding from venture capitalists such as Draper Fisher Jurvetson. As a private company, it is responsible primarily to its investors rather than to the scientific community. So it was no surprise that in announcing its success in building an adiabatic quantum computer, D-Wave focused on commercial applications rather than scientific details. While venture capitalists were impressed by the announcement, treating the company to another round of funding, scientists were less excited. The press release provided no device specifications that would allow the scientific accuracy of its claims to be assessed. It seemed possible that the computer was simply finding solutions by being cooled down to its ground state, a fairly dull and not-so-quantum-mechanical process, rather than performing the more subtle adiabatic procedure described above. When D-Wave neglected to supply any concrete evidence that the device was actually performing a quantum computation, even the most charitable scientific observers simply assumed that its scientists didn’t know whether it was or not. (See “Desultory D‑Wave”) Less charitable observers uttered words I cannot report in this publication. For my part, I was conflicted. I would really like to know whether adiabatic quantum computation works. Even if this approach can’t solve the very hardest problems, if D-Wave’s system could perform a well-defined demonstration of adiabatic quantum computation in some simple instances, that would be a validation of Kaminsky’s and my design. As matters stood, however, D-Wave seemed to be muddying the quantum well for money.
Last fall, the waters became clearer. D‑Wave’s chief theoretician, Mohammad Amin, and its chief experimentalist, Andrew Berkley, visited the quantum-computing community at MIT. They discussed the scientific issues frankly. No, they admitted, they couldn’t prove that what they were doing was true adiabatic quantum computation–but it looked as if it probably was. How could they answer the question conclusively?