[oldgradstudent]

In what sense the result describes a computation? And it what sense the device describes is a computer?

The only thing it can do faster is to run itself.

It's as if I called a Boeing airliner an aerodynamic computer and each flight a computaion.

[krastanov]

They are programmable, that makes them drastically more intetesting than a single-parameter aerodynamic analog computer like the Boeing, from the point of view of Computational Complexity. It does not make them "useful" yet, but it is a big milestone. See https://scottaaronson.blog/?p=5460

[oldgradstudent]

Programmable, or parametrizable?

In any test flight there are plenty of parameters set for the flight. Are aerodynamic computers programmable?

But in any case, the mere existence of "quantum supremacy" research is a clear indication these are pointless contraptions.

[krastanov]

Could you elaborate on your last paragraph? Proof of concepts that can not solve "useful" problems seem like an incredibly important milestones to me? Or is your frustration that they are occasionally presented by overly enthusiastic engineers as more than "useless" technology demonstrators (with this frustration I would agree).

The second half of the article I shared covers the parametrization question you raised.

[oldgradstudent]

> Could you elaborate on your last paragraph?

A successful technology does not need to show theoretical supremacy. Its proponents can simply show the useful services and products it makes.

> Proof of concepts that can not solve "useful" problems seem like an incredibly important milestones to me?

Sure, just don't claim supremacy until you can back it up with an actually reasonable definition of computation.

> Or is your frustration that they are occasionally presented by overly enthusiastic engineers as more than "useless" technology demonstrators (with this frustration I would agree).

It is hard to find any news about quantum computing that is not way over exaggerated.

> The second half of the article I shared covers the parametrization question you raised.

His argument may apply to the teapot example, but for, say, a test flight to fine tune the auto pilot behavior, there are plenty of inputs that can be set. I'd argue the aerodynamic computer supremacy example still stands.

[krastanov]

To your last point: I think I disagree, because your analog computer (like all analog computers) addresses one fixed-size instance of a problem, not the scalable family of problems. You can not make a computational complexity big-O statement about your analog computer, but you can do that about boson sampling and random circuit sampling. This is a crucial milestone in the creation of a physical realization of a computational device.

In your first paragraph, while I am sympathetic with your annoyance (pardon my imprecise choice of words), I disagree there too. Mostly because I can imagine the same argument used against Babbage, Lovelace, and Turing, who did their theory (and failed experiment) work decades to a century before a scalable classical computer existed.