Hacker News Reader

Comment by reikonomusha | Hacker News Reader

Miniaturizing quantum computers is the absolute last priority issue in the field right now.

Making quantum computers “scalable” means overcoming challenges in qubit coherence times and operational fidelities, as well as manufacturing yield.

Putting a quantum computer—that doesn’t do anything remotely useful—on a rack, or in a glass case, or whatever is, in my not so humble opinion, just a gimmick.

While I think the miniaturization is a gimmick, it’s nonetheless nice to see another quantum computer built with respectable fidelities. It’s not good enough to do anything other than further science experiments, but it does show the scientists’ care for the details and the design.

View on HN

This seems a bit reductive.

Sure. There are larger challenges but size, specialization and cost are three things that prevent adoption of this technology at all universities and some companies.

Having more adoption even if its "useless" right now means more people looking for useful things to do.

I think this is a key misunderstanding.

The issue is not that we don't know what useful things to do with a QC. The issue is the quantum computers that exist today are not able to do those things.

"Adopting" a machine that can't do anything you couldn't do on your laptop won't help. The challenge for QC to become useful is to scale up and provide working error correction.

I agree completely, I would love quantum computers to be in a place where all universities and most companies can do useful stuff with them, but they aren't and these companies which pretend that they are seem like scams.

Genuinely curious then. When you say that it can't do anything that my laptop can't do. Is that based on simulation of actual quantum methods with all of the quirks or simply traditional algorithms?

If it's not truly simulation, there seems to me a ton of value in having even a little bit of access to something like this.

It's both. You can easily simulate 11 qbits.

There's a commercial product for classical simulation of small QPUs [0]. With dedicated hardware handles up to 41 qubits. I'm not affiliated with the company or endorsing their products, just pointing out that they exist.

[0] https://atos.net/en/solutions/quantum-learning-machine

You can also do this with purely free and open source software like [0].

[0] https://github.com/quil-lang/qvm

The wavefunction of an n qubit system is representable by a 2^n vector of complex numbers [1]. Using single precision floats [2], a 10 qubit wavefunction is 8 KB, 20 qubit = 8 MB, 30 qubit = 8 GB and 40 qubit = 8 TB.

Hence a run-of-the-mill laptop can easily simulate a perfect 30 qubit computer, provided you are using a sufficiently performant simulator (such as https://github.com/qulacs/qulacs ).

[1] This is actually a pessimistic upper bound for the classical simulation because current quantum hardware is not fully connected and not fully coherent.

Most problems of interest (optimization, chemistry etc) also do not ever generate fully entangled wavefunctions and so can be simulated with significantly less resources.

So for any applications (beyond simulating quantum advantage experiments built specifically to make classical simulation hard), classical computers are crushing quantum computers.

[2] Because QM is a linear theory, numerical precision isnt an issue, as it would be in a chaotic classical simulation of eg fluid dynamics.

The rabbit hole goes deep in this. Form factor isn’t limiting adoption. Bad performance and the fact you need a team of physicists to calibrate the machine daily are two simple examples limiting adoption more than size. I’m not saying miniaturization isn’t useful, I’m saying it’s last on the list to care about.

> I think the miniaturization is a gimmick

I suppose you can make it useful by inserting more q-bits into the expansion slots (?)

That's not how it works.

Two 11 qbit quantum computers are not like one 22 bit quantum computer.

Quantum computers become powerful when you can entangle many qbits. That's also the challenge, because building such a thing is hard.

Reply on news.ycombinator.com