Pages

May 16, 2011

Scott Aaronson comments on the Dwave Nature paper that provides evidence of quantum annealing

Scott Aaronson has commented about the Dwave paper that was published in Nature. (Quantum annealing with manufactured spins)

In the new work, they apply an annealing operation to eight coupled qubits arranged in a 1D chain, then plot the probability of a particular basis state as a function of time, by running the experiment over and over and stopping it at various intermediate points. They then look at the dependence of the probability-versus-time curve on a third parameter, the temperature, and claim that they can explain the curve’s temperature dependence by a numerical simulation that assumes quantum mechanics, but not by one that assumes classical simulated annealing.



To be clear, an eight-qubit spin chain with a quantum-mechanical temperature dependence is still a very long way from anything commercially useful (and it’s notable that, now that D-Wave has happily joined the ruling-out-the-null-hypothesis club, we’re down from 128 qubits back to 8). This paper also makes no claims to demonstrate entanglement, which is almost certainly necessary for any interesting quantum speedup, and which has been verified in other superconducting qubit experiments (e.g., the Schoelkopf Lab‘s at Yale), but as far as I know still not in D-Wave’s.

No, I don’t have any regrets about pouring cold water on D-Wave’s previous announcements, because as far as I can tell, I was right! For years, D-Wave trumpeted “quantum computing demonstrations” that didn’t demonstrate anything of the kind; tried the research community’s patience with hype and irrelevant side claims; and persistently dodged the central question of how it knew it was doing quantum computing rather than classical simulated annealing.

D-Wave has done a real experiment that deserves the careful scrutiny it will receive.

I’m a theoretical computer scientist, not a physicist (much less an experimentalist). So in previous posts, the only reason I even presumed to comment on experimental matters is that D-Wave made it easy for me! My “expert analysis” mostly just consisted of pointing out, over and over, that D-Wave hadn’t yet brought the QEDB (Quantum-Effect-Demonstrating Beef)—and that, until they did so, there seemed to be little reason even to discuss the other issues that D-Wave’s marketing materials and the press were both spending 95% of their time on. Now that a slice of QEDB (or something that looks like one, anyway) is on the table, I think there’s at least as much need as ever for critical evaluation of D-Wave’s claims from the quantum computing research community, but I no longer see Shtetl-Optimized filling that need. So I hereby announce my retirement as Chief D-Wave Skeptic, a job that I never wanted in the first place.

Suz Gildert (Dwave) comment - I’m really confused as to where this myth about D-Wave’s qubits being noisy comes from. The qubits have amongst the lowest values for 1/f flux noise ever reported in the literature.

See for example: Experimental Demonstration of a Robust and Scalable Flux Qubit R. Harris et al. – Physical Review B 81, 134510 (2010)

Quote:

“While noise is not the central focus of this article, we nonetheless present experimental evidence that, despite its physical size and relative complexity, the observed flux noise in this flux qubit is comparable to the quietest such devices reported upon in the literature to date.”

Values close to 1 microPhi0 per sqrt Hz at 1Hz are reported in this study. That’s a pretty low noise level for RF-SQUID based flux qubits.


If you liked this article, please give it a quick review on ycombinator or StumbleUpon. Thanks
blog comments powered by Disqus