>>12767926While I agree with his general sentiment that the popsci news needs to be toned down with regards to quantum computing, the details don't pan out.
1. Google's supremacy experiment is not a "teapot experiment" in the sense he's trying to bring across. The Sycamore chip is not specialized to do random circuit sampling. It seems like his teapot test was thought of when D-Wave was spewing it's nonsense, but nowadays the field is more or less sure that D-Wave is garbage and Google's (and IBM's and Honeywell's and etc.) programmable chip is much less relevant to the teapot analogy.
2. He's talking about whether "quantum computation is currently, in practice, better than classical computation" but there's no person in the field of QC who believes that statement. He's essentially arguing against a strawman, I'm afraid to say. We know that classical computers will perform better in any useful task, where "useful" can be defined incredibly broadly. Anyone suggesting otherwise is delusional and shouldn't be taken seriously.
3. On the other hand, if he wished to compare the theoretical capabilities of quantum vs. classical, then it's no question that there are things that quantum computers can do that classical computers, at least to the best of our knowledge currently, cannot. Quantum computation as a framework generalizes classical computation, and so anything that a classical Turing machine can do in polynomial time, so can a quantum Turing machine. Meanwhile we know of some examples (e.g. order finding for integer factoring) where the best-known classical algorithms are superpolynomial, yet we have poly-time quantum algorithms. This doesn't prove that P != BQP necessarily, but it provides fairly strong evidence.
4. Finally, he remarks on boson sampling and claims that it falls under the "not interesting" category. But this is patently false, because it's literally just computing the permanent of a unitary matrix, a calculation as old as linear algebra itself.