Ars Technica has the article Here’s what the people who claimed Google’s quantum supremacy have to say about it. I have been reading various headlnes about this, but here is the first article I have actually read.
Google’s argument for quantum supremacy focused on the claim that a simulation of its processor’s behavior would take 10,000 years on a state-of-the-art supercomputer. But IBM noted that Google’s argument was based in part on memory starvation, and supercomputers have hard disks that can hold temporary data during the computations. If that disk space is factored in, IBM argues, the calculation could take as little as 2.5 days. At a couple of minutes, the quantum processor beats that handily, but there’s still the chance that algorithm optimizations will cut the margin considerably.
I choose this excerpt, not to denigrate Google’s achievement, but to put some perspective on it for people not versed in computers.
In my day, the hot topic was massively parallel computers. My group at Digital Equipment Corporation (DEC) did some work in the field, so I came to understand the limitations of new technology that is in search of a problem that it can solve. I even remember when lasers were called “a solution in search of a problem”. The problems that they can solve do eventually get found, but it takes a while.
Gabriel Bischoff and I have a patent for some software that took advantage of parallel computers (that were not massively parallel) Method and apparatus for circuit simulation using parallel processors including memory arrangements and matrix decomposition synchronization. This patent is an extension of circuit simulation techniques to take advantage of parallel processing computers. Digital Equipment Corporation’s version of SPICE was the vehicle upon which these techniques were demonstrated. What we found is that as we applied novel techniques to solve the problem on a parallel computer, we realized that some of those techniques could be used on non-parallel computers to speed up the computation.