No; they do not actually factor any 1024 or 2048 bit numbers. The largest is 17 bits. Although they point out that representing the problem can be done in a quadratic size of the input, the datapoints in the paper don't give any reason to believe that the time to finding the factors won't just be exponential in the input.
Also, no, this is not a record for Quantum Computing; as this is a DWave Quantum Annealing machine which needs to be evaluated by a different standard.
This seems to be more interesting if you look at it from the perspective "How can I treat prime factorization as a optimization problem" rather than "Does it look like Quantum Annealing Machines can factor RSA in the near term".
The record for Shor's algorithm is still 21 = 7 x 3
You can instead do "big" numbers like in this article with machines that can't run Shor's algorithm, but there is no good reason to imagine these would ever be competitive with a conventional computer like your laptop or phone.
Shor's algorithm (and similar algorithms) are definitively faster, if you can make a quantum computer to run them. That's the hard part, and as you observe the result is no clear progress in almost a decade.
Quadratic in size of input is just the multiplication table of the radix units in some base (in their case binary).
It's interesting and I think promising approach to factoring is to use annealing over such a multiplication table trying to resolve the intermediate radix unit products and carries. It's a good way to map the problem to a solution space that is obviously a solution landscape that can be explored.
It's true that the number of variables they end up with is still exponential in the bits of input, but then again, that's not really a problem for true quantum computers.
It definitely seems factorization is closer to being publicly broken.
No; they do not actually factor any 1024 or 2048 bit numbers. The largest is 17 bits. Although they point out that representing the problem can be done in a quadratic size of the input, the datapoints in the paper don't give any reason to believe that the time to finding the factors won't just be exponential in the input.
Also, no, this is not a record for Quantum Computing; as this is a DWave Quantum Annealing machine which needs to be evaluated by a different standard.
This seems to be more interesting if you look at it from the perspective "How can I treat prime factorization as a optimization problem" rather than "Does it look like Quantum Annealing Machines can factor RSA in the near term".