[1] Many colleagues told me they thought experimental BosonSampling was a dead end, because of photon losses and the staggering difficulty of synchronizing 50-100 single-photon sources. They said that a convincing demonstration of quantum supremacy would have to await the arrival of quantum fault-tolerance—or at any rate, some hardware platform more robust than photonics. I always agreed that they might be right. (Update 12/5 2020)
[2] But other skeptics are current and former members of the Google team, including Sergio Boixo and John Martinis! And—pause to enjoy the irony—Gil has effectively teamed up with the Google folks on questioning the new claim. Another central figure in the vetting effort—one from whom I’ve learned much of what I know about the relevant issues over the last week—is Dutch quantum optics professor and frequent Shtetl-Optimized commenter Jelmer Renema. (December 16th, 2020)
[3] It is already clear, that in view of Kindler and mine 2014 results, the claims of achieving in 200 seconds samples that require millions of computing years on current supercomputers are unfounded. (January 6, 2021)
[4] Without further ado, why might the new experiment, impressive though it was, be efficiently simulable classically? A central reason for concern is photon loss: as Chaoyang Lu has now explicitly confirmed (it was implicit in the paper), up to ~70% of the photons get lost on their way through the beamsplitter network, leaving only ~30% to be detected. At least with “Fock state” BosonSampling—i.e., the original kind, the kind with single-photon inputs that Alex Arkhipov and I proposed in 2011—it seems likely to me that such a loss rate would be fatal for quantum supremacy. (December 16th, 2020)
[5] Part of me feels guilty that, as one of reviewers on the Science paper—albeit, one stressed and harried by kids and covid—it’s now clear that I didn’t exercise the amount of diligence that I could have, in searching for ways to kill the new supremacy claim. But another part of me feels that, with quantum supremacy claims, much like with proposals for new cryptographic codes, vetting can’t be the responsibility of one or two reviewers. Instead, provided the claim is serious—as this one obviously is—the only thing to do is to get the paper out, so that the entire community can then work to knock it down. Communication between authors and skeptics is also a hell of a lot faster when it doesn’t need to go through a journal’s editorial system. (December 16th, 2020)