In what sense is quantum computing a science?
cognitivemedium.com
cognitivemedium.com
Input A yields outcome B.
We get that understanding this way:
I love that whole series of lectures. They are his Cornell introduction to physics lectures and are a gold mine of basic scientific type thinking.
So far, science has not increased our knowledge of truth. That's damn tough to come by.
What it has done is very significantly expanded our understanding of the world. Our theories are far more predictive in far more contexts than ever!
But, that is it! Monkey see, monkey do.
Engineering is all about applying that understanding to solve problems in robust, efficient, safe ways.
Science uses mathematics to work out the implications of precise prediction-generators (often called laws) and check them for logical consistency, but mathematics itself has nothing to say about any world outside of the one generated by its axioms.
Anyway, bringing this back to QC. It is believed that algorithms exist that can be computed by a QC that cannot feasibly be done by a classical computer, so QC also meets the ”deductive mathematical definition with applicable characteristics” bar as well.
Anyway, let me be the one bringing the discussion back to quantum computing this time. I don't think anybody in the world who could argue that study of entanglement - think Bell's theorem and CHSH inequality - isn't science. In fact, the 2022 Noble Prize for Physics was awarded to people who made foundational contributions on that field. I see Quantum Computing as a direct spiritual successor to that study.
A proper definition of science would be able to consider a study X and classify it as science or not science using only intrinsic properties of the study, not an extrinsic property like a known connection to the empirical world.
https://en.wikipedia.org/wiki/Umbral_calculus
(you can see in that Wikipedia article the umbral calculus was originally invented and used because it seemed to work, long before anyone found a set of definitions to justify it to modern standards)
The axiom of choice is accepted because it's useful and required for certain results to hold that mathematicians aren't willing to give up, etc. If ZFC were to be somehow found to be inconsistent it would probably be patched up by altering the axioms rather than wholesale tossing out all the theorems.
I guess what I mean is that mathematics uses deduction in proofs but that's really not the end (or the beginning) of the story.
As a practical matter a great deal of mathematics was and is done prior to being axiomatized.
Computer science isn’t really a science either. It should be classified as a branch of mathematics.
Computer Science Is Not a Science - https://cacm.acm.org/magazines/2013/1/158757-computer-scienc...
- finding and implementing in hardware an "embedding" which solves certain classes of problems efficiently, especially for which there is no known "efficient" process otherwise.
- finding an elegant way of describing those classes of problems and
- routes for identifying them.
I think [0] Schwartz-Zippel as applied to Polynomial Identity Testing is an instructive example where using noise as a computational resource improves efficiency.
[0] https://en.wikipedia.org/wiki/Schwartz%E2%80%93Zippel_lemma
Ignoring the substantial engineering challenges, if QC doesn't work, we would learn something very interesting about our world in how it breaks, because any break would represent a deviation between the predictions of QM and reality, something we desperately need to get to the next level of theory.
Contrary to what some people might think, physicists would be giddy over a clean, lab-reproducible, engineering-type break in QM. They are starved for data right now. To be able to push a button and reproduce it is an almost impossibly beautiful dream. Almost for that reason, I expect our problems in QC to remain firmly in the engineering domain, and that no matter how many qubits we manage to deploy they will obstinately perfectly conform to QM theory.
Experts are clearly more optimistic, especially Scott Aaronson who is often mentioned on HN, but I can't get past a couple of simplistic, non-mathematical objections.
One is that it seems already like quantum computing is on a different trajectory than classical computing. If it was an analogous engineering problem, why haven't we already gotten much farther?
The other is that it just seems like a matter of symmetry. Quantum computing, even if it's not all powerful, seems like a "cheat code" for reality, and an inability to practically exploit it seems to me like balancing it out. I believe in a vague heuristic that reality has no "thread" that can be pulled to unravel everything.
I guess the counter to these vague feelings are that you could've said the same stuff about nuclear fission/fusion and any technology based on quantum theory.
So I'm not that confident in my opinion, but I'm still seeking an understandable reason why one should be optimistic. Nothing I've read by experts translates their feelings to my understanding.
The sort of situation that seems intuitive to me, I don't see why it would result in any earthshattering violation of theory either, as we'd just approach an asymptote forever.
P.S. I have a similar attitude towards practical fusion power for somewhat similar reasons, so that's evidence it's an attitude problem more than anything.
If you think about it even a little, it is almost axiomatic to the intelligence/security world that a machine that could break much known encryption in human relevant timeframes would obviously be being developed under a shroud of secrecy at the bleeding edge. It is far too important to be able to manage the implications of any significant breakthroughs potentially years before public release.
QC is not magic and there are effective counter strategies. But standards take time to implement and the well being of society demands a that the impacts of QC be mitigated through control of the tech and the information around it. Fortunately, the tech required so far keeps access to the computing resource somewhat restricted, and organizations that cooperate with governments control it so far.
For most people, the important thing to keep in mind is that everything now shielded by many forms of encryption will be trivially readable in the foreseeable future. Truly sensitive data stored long term should be moved to Q hard algorithms and old versions destroyed as possible (easier said than done in many cases)
I'm unconvinced of this. Any government agency (say the NSA) is a tiny subset of the whole world, and can't be expected to outdo it.
Especially if they have some disadvantages like, say, not being able to hire people who've ever smoked pot.
>It is far too important to be able to manage the implications of any significant breakthroughs
But that's what keeping breakthrough tech secret would do - prevent managing the implications. Keeping it secret means not sharing it with allies. That means if someone else discovers the same thing or steals it via espionage, now your adversary has it and your allies don't.
That is how the NSC/MIC works, it’s mostly outsourced. The government per se does very little research with some notable exceptions.
They have some things that work the difficulty down below exponential, which seems reasonable, but it's still a pretty stiff poly factor (by engineering standards; it's actually pretty impressive mathematically but right now even having to add a small constant factor of qubits would be a kick in the teeth from an engineering perspective, and of course it's worse than that). I'm fairly skeptical myself. However, it's the sort I consider "true skepticism"; it isn't a certainty it will fail rhetorically disguised as uncertainty, I'm really not certain. If someone produces a working, useful QC device, then it is working and useful regardless of my skepticism, and I will celebrate them for all the more for the fact I thought it was really hard.
"Quantum computing, even if it's not all powerful, seems like a "cheat code" for reality, and an inability to practically exploit it seems to me like balancing it out"
IMHO, that's more the hype than the reality. The reality is that we don't have that many QC algorithms in general, and we have even fewer algorithms that are clearly advantageous over classical computing. If we had a perfect QC device with billions of qubits in hand right now and could mass produce them cheaply, at the moment I think the evidence is that you'd be looking at something more like a GPU situation where they get added on to conventional machines for certain problems than a complete revolution. And to be honest a lot of people wouldn't. Graphics cards give you awesome gaming; a QC attachment will solve specific problems that we just don't have everywhere, and using QC to do something a conventional computer can already do is going to be as silly as treating your GPU as a CPU.
The primitives that QC offer us are really weird. It's completely different, but in terms of understanding the difficult, imagine trying to write computation directly in terms of Conway's Game of Life, because we for some reason have some sort of massive accelerator that can run it quadrillions of times faster than we can simulate it otherwise, and trying to get it compute anything useful in "native" Life, e.g., if you do what we would really do in this sort of circumstance and lay a Turing Machine over it you're losing the "advantages" of Life in the process. QC is actually not quite that bad in my opinion, but I think it at least gives the flavor of the difficulties of QC. And also simulates the fact that you can'd do what we usually do in the discrete domain and just lay a conventional interpreter over whatever the problem is and work in terms of that interpreter, because you have to have the algorithm "stay quantum" for the QC to be of any use.