From vacuum tubes to qubits – is quantum computing destined to repeat history?

The early 1940s saw the first vacuum tube computers put to work solving problems beyond the scope of their human counterparts. These massive machines were complex, specific, and generally unreliable. In many respects, today's quantum systems bear remarkable similarities to early vacuum tube computers in that they're also …

[nautica]

'Quantum Computing' IS repeating history.

"...is quantum computing destined to repeat history?"

Yes; repeating scientific history as provided by cold fusion, JB Rhine's attempts at legitimizing parapsychology, the use of nuclear fusion for green power production, Prosper-René Blondlot's N-Rays, and further--and continuing--validation of Wolfgang Pauli's aphorism, "That concept is so bad it's not even wrong.".

Yes, quantum computing is not only destined to repeat history, but is.

[MrBanana]

A solution without a problem

So, quantum computing, yeah! it will make breaking crypto problems more easy, apparently yeah! Anything else? Climate change, world hunger - hmm. Will it fit on my desktop, probably not. Will it predict who will win Strictly, gonna say no.

[Anonymous Coward]

Re: A solution without a problem

25 years ago (when I was tangentially involved in such things) one of the suggested applications of a quantum computer was to produce better simulations of quantum-mechanical systems. Which sounds rather recursive, but simulating exactly how a drug binds to a receptor could be of interest, for example. And drug companies tend to have lots of lovely cash to invest in such things.

[nautica]

Re: A solution without a problem

...not only rather recursive, but definitely recursive, a tautology ("Quantum computers can solve those problems that, um, only a quantum computer can solve."(!)). The very valid argument against quantum computers is the same as that which applies to any pseudoscience: the concept is not falsifiable. This is precisely what is wrong with the statements--and very hard work--of some extremely sincere people regarding such things as fusion power to solve all our energy problems ("Oh, so what you're proposing is that we simply build a miniature sun, right here on earth to solve all our problems ‽ Sounds good to me; we'll get right on it. Oh, and I'll only need several millions of dollars to get started, and to continue. Gee, thanks; I knew you'd understand..." ); and Superstring Theory, which would only require more energy than what's available in the observable universe to prove...or disprove.

Ask anyone who professes to be a Q-C 'expert' two simple questions: (1) In terms, which any Electrical and Computer Engineer can understand, how does one build a Quantum Computer?, and (2) demonstrate (same conditions) how one programs a Quantum Computer. It has been proven time and time again that your answer(s) will consist of nothing more than what can be charitably described as 'arm-waving' and 'tap-dancing'.

The Quantum Computer is a problem looking for a problem; which, unfortunately, is going to suck a lot of resources away from valid, tractable problems and efforts which definitely need addressing and for which there exists a reasonable chance of solving. Quantum Computing meets none of these criteria. Simply start with the "...valid problem..." point. One needs go no further.

[Francis King]

Re: A solution without a problem

Try this: 64-bit processors can solve those problems that only a 64-bit processor can solve. Why can only 64-bit processors solve those problems? Because a 32-bit processor cannot address enough memory. It may be a bit of a tautology, but not really recursive.

Some problems will require a quantum computer.

I am not a quantum computer expert, but here goes: we create electronic pockets which contain some electrons, with a given spin or something. They are very cold, and very close together, so the electrons interact, unlike a digital circuit where we hope they won't interact. Hence instead of the classical N separate bits, we have in effect 2^N bits. As N goes north of 80 bits, a digital computer can't keep up. So, that's the arm waving done, but it's hard to avoid arm waving if you also want something that 'any Electrical and Computer Engineer can understand'. Tap dancing is extra.

The person who invented this concept was Richard Feynman. Unlike Elon Musk, Richard Feynman knew what he was about, and it would take a brave or arrogant man to bet against him.

[Michael Wojcik]

Re: A solution without a problem

You really have no idea what you're talking about, do you?

The broad strokes of "how does one build a Quantum Computer" are easily understood, for a variety of designs, by educated people; there are any number of articles available that explain how the major architectures work, such as this one. Josephson-junction, ion-trap, and laser-atom designs are all working in multiple labs, and research on NMR and photonic designs continues. Obviously the technical details are understood primarily by people actually working in the field, as with any other technical field. Your "conditions" are bogus; they're equivalent to "explain how to build a modern automobile engine, in terms which any mechanical engineer can understand". Few mechanical engineers are experts in flame chemistry or CANBUS or other aspects of automotive engineering, because they're not automotive engineers. It's a different fucking discipline.

As for "how one programs a quantum computer": A number of algorithms in BQP are well-documented, and many are quite straightforward. Actual details of programming are going to depend on the actual machine and its software, just as for any other programming task.

Constructing incoherent decision procedures isn't a persuasive argument; it just demonstrates you're not willing to argue in good faith.

[EricB123]

Re: A solution without a problem

"And drug companies tend to have lots of lovely cash to invest in such things."

You mean instead of stock buy-backs?

[Michael Wojcik]

Re: A solution without a problem

Quantum simulation is likely the best use of QC in the foreseeable future. There are potential applications for other problems in BQP, but economically-practical use cases are hard to find.

[StrangerHereMyself]

Applicable

Just like early computers were seen as advanced calculators until their universal applicability was discovered the same will (or could) happen with quantum computers. The early versions will only be used for a select class of computing problems (routing, chemistry) until one day someone discovers you can compute millions of calculations in parallel with them though superposition, which previously had to be done sequentially.

We could therefore see future VR simulations becoming indistinguishable from real-life if the necessary computer-brain interface becomes available.

[Doctor Syntax]

Re: Applicable

"if"

Such a trivial word. However I look forward to Musk taking the lead in testing this concept.

[Michael Wojcik]

Re: Applicable

Sigh. That is not how quantum computers work.

[Phones Sheridan]

Quantum Computing? Will file it along with Fusion and IPv6 in the drawer marked “no meaningful progress in my lifetime”

[Pascal Monett]

You're not dead yet.

[Paul Crawford]

Maybe they are, but the box is not open yet?

[Anonymous Coward]

One can always cut the quantum Gordian knot, by chucking the box in a compactor

[Phones Sheridan]

It’s an active work in progress! Chances of success 100%

[C R Mudgeon]

And yet...

Until a very few years ago, I had AI filed in the "no meaningful progress[1] in my lifetime” drawer too.

I'm reminded of Arthur C. Clarke's comment: "When a distinguished but elderly scientist[2] states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong."

[1] I have grave doubts that AI is proving to be a good idea on balance, but the technical progress has been remarkable.

[2] I'm not claiming any personal distinction, or to have been a scientist; just a long, recently ended career in non-quantum computing. The "elderly" part? Yeah, that's approaching far too quickly.

[Danny 14]

you can prise my quadruple blocks of /29 out of my code dead fingers.

[vtcodger]

A Promising Technology

The thing about 1950s computers is that they quite clearly worked. And they produced useful results. No one doubted that they worked. The issue back then was that it wasn't clear that digital computers would be a cost-effective solution to most real world problems

Despite huge investments and massive amounts of hyperbole it's not clear that ANY of the current crop of Quantum computers actually work. ... Or that the next generation will work. --- Or that their great, great, ever so great offspring won't be a really promising technology just as soon as the techies conquer the last few practical difficulties -- a state that can probably endure damn near forever.

[Michael Wojcik]

Re: A Promising Technology

it's not clear that ANY of the current crop of Quantum computers actually work

That's incorrect. The quantum-circuit-simulation tests, for example, demonstrate pretty conclusively that QCs work. That is, Google and others have demonstrated QC machines which use quantum algorithms to solve problems using fewer operations than any known classical algorithm would use. What's not clear is whether they've actually achieved quantum advantage, since it's really hard to prove that there isn't a better classical algorithm that wouldn't bring the classical time into the realm of feasibility.

And what's even less clear is whether and when we'll have QCs that offer any sort of economic advantage for any plausible real-world problems; and when that happens, how common such problems will be. It still appears we're a long way away from the quantum cryptocalypse, for example, because being able to economically break a lot of (non-post-quantum) asymmetric-cryptography keys using Shor's algorithm¹ would take big, cheap, fast² QCs, not just usable ones.

¹Or Regev's recent improvement, assuming Regev's assumption of the smoothness conjecture holds.

²"Fast" because "fewer operations" doesn't necessarily mean "done by lunchtime".

[amanfromMars 1]

The New Masters of the Universe ....

And whilst every man and his dog wait for quantum computer hardware manufacturers to get their act together and present a viable machine, do quantum computer programmers and systems developers hone and perfect the instructions sets that deliver what should/could be needed.

[Doctor Syntax]

I'm a bit concerned about agreeing with someone from Gartner.

What's always puzzled me about quantum computers - well, one of the things - is this reliance on noise reduction. It sounds a bit like "Here are all the possible answers. All you have to do is work out which is right.".

[C R Mudgeon]

"I'm a bit concerned about agreeing with someone from Gartner."

Now that, I'll go along with.

[Michael Wojcik]

Again, that's not how QCs work. But error correction, which is essentially what "noise reduction" means in this context, is indeed still a big problem.

[Anonymous Coward]

Don't put all your boffins in one basket.

There are a number of unicorns running to be the next frontier in computing. Among them Quantum Computing is the only unicorns with wings. Still, I wouldn't be surprised to see optical/analog/3d-nerve-fiber/something-else computing be first to the mass production gate sooner than the QC horse clears the runway. I think AI is going to soon run into a brick wall on making a profit because of it's energy and resource demands. That is a huge incentive to develop something new.

John Von Neumann's brain ran on less than 20 watts. Yeah so his brain couldn't run Grover's algorithm. It wasn't a deal breaker.

[Michael Wojcik]

Re: Don't put all your boffins in one basket.

Any Turing-complete machine with sufficient resources can run Grover's algorithm. QCs can be simulated on non-quantum computers.

Algorithms in BQP aren't magic; they're just amenable to computation with a lower time (really number-of-operations) complexity when using qubits. If you have to simulate the qubits, the complexity goes back up, but it's still computable.

[kuiash]

Cost/Performance

Assume that there is a class of algorithms that are "well designed" of Q machines and, conversely, not well dealt with on transistor machines.

Assume that there are companies, with deep pockets, that already spend a gazillion dollars on conventional machines.

If they can build (and run) a Q machine for half the cost and twice the speed as a transistor machine then it's a no-brainer for them.

Some of these guys are running crazy algos for crazy reasons.

Do I want one? Nah. I think putting the required cooling in my laptop or phone is... "problematic"!

[Doctor Syntax]

Re: Cost/Performance

That's a lot of heavy lifting for two words. Which two am I looking at?

[Michael Wojcik]

Re: Cost/Performance

Assume that there is a class of algorithms that are "well designed" of Q machines and, conversely, not well dealt with on transistor machines.

If P≠NP, then this assumption is almost certainly true. Unless the complexity hierarchy collapses, BQP is a thing.

[DJO]

Promising or overhyped?

Considering how long useful quantum computing has just been a few years away it is beginning to get the fusion levels of hubris in that it's always "just over the horizon".

There are significant issues with coherence which (currently) push the cost/benefit ratio to levels hard to justify other than in QC research.

As for the early adopter advantage, computing history shows that is not always a good thing, look at J. Lyons and Co. in the UK - early pioneers but once they had competitors they were stuck with an outdated technology.

I may be wrong (it has happened once or twice) but I suspect the future of QC is firmly in obscure niche applications, for the massive majority of general purpose computing QC offers no advantage. I also suspect the code breaking possibilities of QC are based on a misunderstanding of quantum physics and will not turn out to be a serious threat to (well constructed) cryptology.

[Michael Wojcik]

There is no "quantum only" problem

not whether we can go faster, whether we can actually solve new classes of problems

Argh. There are no such classes of problems. Quantum computers can be simulated by classical computers.

What QCs potentially will be able to do is make certain problems feasible at larger sizes. Some of those problems, such as quantum simulation, become intractable for classical computers even at fairly small sizes, so this would be a Nice Thing to Have. But there are no magical quantum-computer-only problems as such, except arguably for engineering ones like "can we build an economically-viable quantum computer?".