3 ms·

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

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.

adamnemecek 8y ago
> 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.
- 6nf 8y ago
  If analog computers are superior why is everything digital
  - adamnemecek 8y ago
    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.
- krastanov 8y ago
  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.
  - adamnemecek 8y ago
    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.
    - krastanov 8y ago
      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.
      - adamnemecek 8y ago
        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.
- artemonster 8y ago
  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.
  - adamnemecek 8y ago
    Ok send me a link. I’ve googled this before but I definitely didn’t find it conclusive enough.

adamnemecek 8y ago

> 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.

6nf 8y ago
If analog computers are superior why is everything digital
- adamnemecek 8y ago
  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.
krastanov 8y ago
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.
- adamnemecek 8y ago
  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.
  - krastanov 8y ago
    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.
    - adamnemecek 8y ago
      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.
artemonster 8y ago
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.
- adamnemecek 8y ago
  Ok send me a link. I’ve googled this before but I definitely didn’t find it conclusive enough.