@forest I know of no actual ICs designed with AI. The routing tools use heuristics with a random seed, so you just press the "compile" button until you hit your timing spec.

I think that one of the things is that the systems are actually simple blocks. You have a transistor. We know a lot about it. It's not like a structure where you can get creative, as you are bounded by a lattice structure.

AI might make routing better. Back in the Alpha days, the 266MHz alpha as one a similar process to the 66 MHz Pentium, but the Alpha was just routed by hand. AI might be able to help with timing critical paths.

@fgrieu I don't see why you're claiming that the integer factored must have some relation with the molecule used. Even more importantly, I don't see why you're saying it's not a general factorization method applicable to any integer of comparable size. The method is general, and for any number, a factorization can be done with the same method, except the number of qubits required will differ.

@fgrieu The factorization of n=21 by Shor's algorithm, was not a valid factorization at all. They used the fact that 21 = 3 x 7 to choose the "base" that they used in the experiment. In John Smolin's paper "pretending to factor numbers with quantum computers", which was later published in Nature with the title "oversimplifying quantum factoring", he and his colleagues showed that for any number, if you know the factors in advance, you can apply Shor's algorithm with only 2 qubits.

n=21 was done in that way! Shor's algorithm has been used "properly" for the factorization of n=15, but that was with a semi-classical version of the quantum Fourier transform, which requires longer coherence times. Shor's algorithm original which would require at least 8 qubits to factor n=15, has never been successfully implemented. Furthermore, to factor RSA numbers you'd need quantum error correction, which has never been demonstrated experimentally to any substantial degree.