JP: Scott Aaronson said that the Orion was as useful as “a roast-beef sandwich.” You obviously feel that is insulting; but wouldn’t you admit that your computer isn’t very useful since it solves problems more slowly than a classical computer?
GR: The purpose of the demo was not to show one-to-one performance superiority over conventional systems. The purpose of it was to do a systems proof of concept and to run commercially relative applications on a quantum computer, which has never even been done before–not even close. This is way above the state of the art. So in terms of the actual time it takes to solve problems, Orion as it currently stands is about 100 times slower than a PC running the best algorithms. If you were an expert, you could define a good algorithm on the Web, spend $1,000 on a PC, and you could beat the system by a factor of 100. So in that sense, Scott’s right, though that’s kind of not the point.
JP: Well, what’s the point, then?
GR: The point is that the demonstration shows a clear path from where we are today into the future. Those future machines will be significantly better.
JP: The plan is to demonstrate a 1,024-qubit machine in 2008?
GR: Yes, by mid 2008. But prior to that, we’re going to have an online system for people to use, for which they can program applications.
JP: That seems implausibly rapid. How will you do it?
GR: Well, there are three things that need to be done.
The first is that the design that you are using for the processor, specifically the input-output systems, need to be scalable, not just in principle but in practice. Most of the proposals that have been put forward for quantum-computing architectures, in fact all of them so far, are not scalable in that sense. In our case, we believe we’ve found a path to real scalability in the hardware. The primary thing that needs to be overcome is this issue of how do you get information into and out of the chip. We think we’ve found a way around that problem.
The second thing is how you build it, and that’s a fabrication issue. Part of the reason why we picked the approach that we picked is that the circuits that we’re using as the basis for these things can be designed, built, and tested using standard semiconductor procedures. So we don’t need to invent any new fabrication technology except for getting the process running in the first place.
The third thing, which is probably the most difficult question to answer, is this: given that we can build it and send information in and out of it, will it in fact continue to operate as a quantum computer? That’s a point that we simply cannot answer at the present time because no one has been able to model systems at that level with any predictive capability whatsoever. It’s too complicated. That’s a question that can only be answered empirically. So our philosophy is, do a new processor every month. Say we have 12 generations per year, something doesn’t appear to be working; we can fix it through iterative redesign.
Smaller design teams can now prototype and deploy faster.