The record for integer factoring on quantum computers was on the order of factoring fifteen into three times five the last time I checked. Can we do three digits now?
Significantly larger numbers than 15 have been factored [1] but not using Shor's algorithm. Shor's algorithm is particularly sensitive to noise/errors in your quantum computer and isn't going to be useful unless we get a properly error corrected machine working. The algorithms used in [1] are considerably less fancy (with worse asymptomatic performance) but are more resilient to noise.
I couldn't quickly find any info, but does this algorithm show the kind of exponential quantum speed up needed to break RSA? Because if it's just slightly faster than the best known classical algorithms, then it's enitely irrelevant to the question of when we need to switch our encryption schemes (even though it may be a significant advancement in the area of quantum algorithms research).
I think its unknown, but my feeling is that the answer is almost certainly no.
These sort of variational algorithms are appealing (to some people) because they're potentially usable on the sort of noisy small quantum computers we have today and in the near term future, but they aren't very fancy. I think in general what you'd expect to get out of them is a sort of Grover's algorithm-like square root speedup.
it's generally believed that the algorithm is somewhere in between ecm and quadratic sieve (so slower by a super-polynomial factor than NFS which is the best classical algorithm)
that paper is factoring with an algorithm that almost certainly isn't polynomial time. That paper is only slightly better than the quantum factoring algorithm of making a quantum computer perform trial division.
And to extend off this comment, there are methods being worked on for building qubits that are intrinsically noise-free and don’t need the exponential number of error correcting operations. When those are available, you’ll see a step function increase in capabilities.
Yeah, for Shor's algorithm to factor an integer of order 2^k you need controlled phase gates with phases roughly order 2^{-k} (very roughly, with some caveats, but lets just say you need some small ones) these very small phase gates are susceptible to even very small errors.
This is a gross oversimplification. For the true version see here
Wot? Science isn't a democracy! The parent refs a preprint from a very reputable author, which has been somewhat peer-reviewed already *
Now, I got to the bottom of page 6 and my maths failed me: I can't follow the expansion, but I expect that the reviewers of Physica A or where ever the gentleman who wrote this sends it off to will be able to check. I do follow the principle of the proof though and it's pretty intuitive to me, for what that's worth.
Anyway, I can't say I give a hoot what the majority or minority think - and nor should anyone else. Read the paper for yourself and make up your mind.
*
The author thanks Al Aho, Dan Boneh, P ́eter Ga ́cs, Zvi Galil, Fred Green, Steve Homer, Leonid Levin, Dick Lipton, Ashwin Maran, Albert Meyer, Ken Regan, Ron Rivest, Peter Shor, Mike Sipser, Les Valiant, and Ben Young for insightful comments. He also thanks Eric Bach for inspiring discus- sions on some of the number theoretic estimates, and we hope to report some further improvements soon [7]. A similar result can be proved for Shor’s algorithm computing Discrete Logarithm, and will be reported later.
I'm not sure that's the right question. It's more, is there a chance at all of anyone figuring it out, and given the enormous scale of the security risk that poses, we should start proactively mitigating those threats. If fusion energy goes from perpetually 10 years away to suddenly here, that's pretty much just a white swan. If quantum computers happen, that's a global security risk before it's a civilizational upgrade.
biggest number factored by a quantum computer isn't the right question. the right question is biggest number factored using a polynomial time algorithm. the answer to that as far as I know of still 15 (although I would be interested in papers that show more progress)
This is one of the things I really resent about QC as a field - there's so much chaff where one paper will say "we can do x" and the reality is that x does not mean what everyone thought that they meant. Number of Qbits is another thing - also what gates are implemented in the devices; how long they can run for etc etc etc.
Application of Shor's algorithm is currently limited by available error correction. Long-lived qubits would eliminate that need and drastically increase capabilities.
I'm not sure that you are correct. I've tried to read https://arxiv.org/abs/2306.10072 in the last day and if my reading is right (I am very stretched by this stuff so I am very happy to be corrected) then no amount of error correction will rescue Shor's - only zero error phase gates. I suspect that a similar story is true for native QML, as quantum memory scales it's just going to get exponentially harder to maintain it.