[reikonomusha]

Why do you think using two-level systems is fundamentally flawed? Even those are analog devices that span a finite dimensional space, but they’re nonetheless dense vector spaces over C^2^n. You still get theoretically infinitely parameterizable operations, though we know some discrete subset of those is sufficient for universal computation.

If one were to work with an infinite dimensional system, you’d still be employing finite truncations of infinite dimensional operators, leading you back to, more or less, a finite dimensional subspace.

Digitization—or rather, discretization—is important, and the reason computers have managed to be so successful. And programming a system like a universal gate-based quantum computer can be done now, with languages like Quil and libraries like pyQuil [0].

[0] http://pyquil.readthedocs.io/en/latest/

[adamnemecek]

Because you lose integration and differentiation in hw. You can also model signals "natively".

[reikonomusha]

You actually get those benefits, differentiation and integration, in the usual classical analog circuits, but, except for a few kinds of analog tricks, they’ve not outperformed their digital counterparts in precision, accuracy, or speed. You can make for neat demos, like wiring up integrators to solve the Lorenz attractor equations to make a nice oscilloscope plot, but the circuits fall short practically for anything more difficult.

[adamnemecek]

I'm well aware. Those that you are talking about are electric though which brings a whole class of issues. I think that photonic might work very well.