Interactive online quantum circuit simulator | Hacker News Reader

Interactive online quantum circuit simulator

strawberryfields.ai

48 pointsghosthamlet20 comments

The company behind this is loaded with the most PhDs I've ever seen in a startup:

https://www.xanadu.ai/about/

Makes sense, it’s hard shit

I get some of the parts but idk what others do. What’s the best resource on this?

This is actually unusual compared with other quantum simulators as it’s a continuous variable quantum computer, not a digital one.

No, this is quite certainly a model/simulator for a digital quantum computer (implemented on top of a harmonic oscillator).

It is becoming a pet peeve of mine, but on most HN posts about quantum computing somebody always mentions some misunderstood statement about continuous variables. Yes, there are many interesting ideas around the use of harmonic oscillators (the older name for the vague and misleading term "continuous variables") to store data, but most of the time it is just one of many ways to implement the same digital computer.

From https://strawberryfields.readthedocs.io/en/latest/quantum_al...

Here we will discuss several continuous-variable (CV) quantum algorithm examples that can be run with Strawberry Fields.

> It is becoming a pet peeve of mine, but on most HN posts about quantum computing somebody always mentions some misunderstood statement about continuous variables.

That someone is me. I’d love to hear what I misunderstand.

> the time it is just one of many ways to implement the same digital computer.

Yeah, what I’m talking about isn’t a digital computer.

Admittedly, I overreacted and I will edit my previous post. The gist of my complaint is that digital vs continuous variables is a false distinction. One is just a system on top of which to implement the other. The useful computation you get out of the quantum oscillator, comes when you implement your digital logic on top of it.

> One is just a system on top of which to implement the other.

We fundamentally disagree on this point. A continuous variable computer is more powerful than a digital one. I know error correction is tricky but I’m still convinced that the analog computer is superior.

If analog computers are superior why is everything digital

Around ww2 analog electric computers were pretty common however electricity suffers from something called rift, basically they get inaccurate over time.

We developed digital computers as they fix the problem with rift.

Photonics don’t suffer for rift so at some point we should revisit analog computing.

This is an incredibly strong statement that virtually nobody in the field agrees with.

1. Continuous variable (in the context of quantum computing) is frequently different from "analog". See bosonic codes for instance.

2. We stopped using classical analog computers because they suffer from noise (which is a fundamental problem, not just a technical issue). Noise is even worse in the case of quantum computing (the claim you make in a sibling comment that some analog systems do not have that problem is wrong/unsupported by any evidence).

3. The main point of disconnect is that noise is a fundamental issue stemming from the laws of physics and computational complexity. There is a large body of evidence that shows that the existence of a scalable analog computer (classical or quantum) will break many structures in computational complexity. A hunch is not enough of an argument to counter those. Presenting a hunch like that is extremely misleading to the people who are trying to get into the field.

4. A hunch is certainly a good thing to research, but do not present it as something of equivalent value to established facts or empirically supported conjectures.

1: imagine I put analog in quotation marks.

2: correct but that was electricity. Light still has that problem but less.

3: let’s get more concrete. Which particular papers do you have in mind? So what is all the continuous variable research about?

> Presenting a hunch like that is extremely misleading to the people who are trying to get into the field.

I’m sure that the number of people who get into the field based on my comment is 0.

To 3: NP-complete Problems and Physical Reality by Scott Aaronson is a pretty good review.

To 2: It is about scaling. The noise might be an order of magnitude smaller, but it does not matter, because this is a constant factor that is overpower by the fact that noise adds up quickly with number of elements/operations, no matter what hardware you use. It is the adding up that is the problem.

To your last comment: you really will be surprised. It is our duty as scientists to balance excitement about the unknown (better analog computers would indeed be awesome) with the desire to speculate about what the solution would be.

Lastly, learning why something is improbable (e.g. a perpetual motion machine) frequently brings amazing insight into the workings of the universe.

I’ve skimmed the paper before. Do you think you could point out the relevant segment? To be honest I’m not particularly fond of Aaronson’s work, it’s a little too...bombastic?

2: it’s all about an error correction protocol. I’m not convinced it’s truly impossible.

what you mean is a machine that can operate on real numbers with arbitrary precision (which is not possible in real world anyways, because of the noise). google that and you would see that they are not "more powerful" than your usual TM. case closed.

Ok send me a link. I’ve googled this before but I definitely didn’t find it conclusive enough.

> on most HN posts about quantum computing somebody always mentions some misunderstood statement about continuous variables.

Well, the D-Wave quantum computer seems to be functionally an analogue machine, so while you're limited to some variety of floating point precision when specifying the values for qubit and connection bias, it's certainly not 0 and 1 when it hits the hardware.

If you are comfortable (mathematically) with the idea of "a gate" and "a qubit", then Wikipedia is surprisingly good at explaining what each new gate does.

Otherwise the standards reference that starts from scratch (prerequisites being linear algebra and intro quantum mechanics) is Quantum Computation and Quantum Information by Nielsen and Chuang. It is a very thick book and it takes a sustained effort to read it, but stackexchange is always eager to help. Preskill's lecture notes are amazing as well.

Regrettably, I do not know of a good book that does not have the prerequisites I mentioned.

Mike and Ike don’t mention for example homodynes afaik.

With classical gates it is possible to make a complete characterization by enumerating the inputs and corresponding outputs (e.g. in table format). Would that be easy/possible here? How big would the table be?

Something like that is possible. A quantum gate is a linear operation, therefore it can be represented as a matrix (it happens to be a unitary matrix).

A gate acting on N qubits would be a square matrix of 2^N rows.

Reply on news.ycombinator.com