Nice write up, although I have a different view of SA. He may be critical of some QC papers, but he also boosted Google's Quantum Supremacy, which was basically a snow job. Despite the paper's claim of solving a "computational task", the quantum gates were unique, device-specific, calibrated, non-standard unitary transformations. A second copy of the device would have had different calibrations and would have reached different results. If we insist that "computational tasks" are device-independent, the study failed.
SA quote: (1) It’s actually really, really hard to invent a model of correlated noise that kills scalable QC—not merely in the weak sense of killing some specific fault-tolerance scheme, but in the strong sense that there’s then a polynomial-time classical simulation of your system. Gil Kalai, an excellent mathematician, has been trying to do this for 15 years, but as far as I can tell has completely failed.
(2) If QC-killing correlated noise existed, and if it were impossible to get around it for fundamental physics reasons (as opposed to it merely being a hard engineering problem), then that itself would be a huge new discovery about physics. It would mean there was something of enormous importance we hadn’t understood about quantum mechanics itself, something that allows a “polynomial-time hidden variable theory.” Nobel Prizes would be in order. I and most of my colleagues would eagerly accept that as an outcome from QC research!
Nice write up, although I have a different view of SA. He may be critical of some QC papers, but he also boosted Google's Quantum Supremacy, which was basically a snow job. Despite the paper's claim of solving a "computational task", the quantum gates were unique, device-specific, calibrated, non-standard unitary transformations. A second copy of the device would have had different calibrations and would have reached different results. If we insist that "computational tasks" are device-independent, the study failed.
SA quote: (1) It’s actually really, really hard to invent a model of correlated noise that kills scalable QC—not merely in the weak sense of killing some specific fault-tolerance scheme, but in the strong sense that there’s then a polynomial-time classical simulation of your system. Gil Kalai, an excellent mathematician, has been trying to do this for 15 years, but as far as I can tell has completely failed.
(2) If QC-killing correlated noise existed, and if it were impossible to get around it for fundamental physics reasons (as opposed to it merely being a hard engineering problem), then that itself would be a huge new discovery about physics. It would mean there was something of enormous importance we hadn’t understood about quantum mechanics itself, something that allows a “polynomial-time hidden variable theory.” Nobel Prizes would be in order. I and most of my colleagues would eagerly accept that as an outcome from QC research!
https://scottaaronson.blog/?p=6813#comment-1943959
Really interesting! loved it.
Thanks for this. I suspect the encryption-breaking FUD will come, therefore I appreciate your feet on the ground thoughts on the subject.