Having worked in quantum computing, I think we'll get fusion first. Essentially the problem is, all of our scalable quantum architectures are really just mathematical models. There's scant research on actually building anything larger than about ten qubits. When you consider fact that error correction means you probably need about 500k physical qubits to do anything interesting (at least!) and as I like to joke even if we followed Moore's law it would take 32 years or so (and I doubt QC would scale so quickly) so not anytime soon.

But that doesn't mean it's "frustrated". There's a lot of progress being made. There's just a long way to go. Quantum algorithms are hard.

Fusion guys have always claimed they're underfunded w.r.t. experiments and in fairness fusion does get kinda hot so I could see it being expensive.

https://commons.wikimedia.org/wiki/File:U.S._historical_fusi...

I think the following viewpoint has been interesting. If we look back at the development of the electronic computer, it's hard to say there was a single point it "delivered". Was it the Atanasoff–Berry Computer, the Collosus, the ENIAC (1940s)? Each of these miss some things that we commonly think of as a modern, generally useful computer. (Lacking programmability, Turing completeness, or stored program representation, respectively.) Do we have to wait all the way until, say, Apple II (1977) was released to consider delivery?

I don't have an answer myself, but it seems to most people, "delivery" doesn't mean "when it's useful", but rather "when it's usable, useful, and generally available for use."

If we use that as a benchmark, quantum computing has definitely hit "usable". It is teetering on "useful". And the beginnings of "generally available for use" have been explored by a few industrial players [0, 1].

[0] Rigetti Computing's Forest: http://forest.rigetti.com/

[1] IBM Research's Quantum Experience: http://www.research.ibm.com/quantum/

The first successful commercial electronic computer was the UNIVAC I. About 18 were delivered to customer sites and operated for years. Earlier machines (ENIAC, SSEC, LEO, Ferranti Mark I) were one-off prototypes. (Two in the case of the Ferranti Mark I.)

I'm not sure about either of these.

The problem with fusion is commercial viability. Building something for billions of dollars that can barely put out more energy than you put in isn't commercially viable. At any rate, fusion seems more like a very tough engineering problem, the physics has been there for a long while.

With quantum computing there are also no real signs that a quantum computer will beat a "classical" computer at anything any time soon. It always was and still is very difficult to get quantum systems to scale and you need a large enough system to make a difference and it may be impossible to do so. Again a lot of this is more of an engineering problem though admittedly the line between physics and engineering can be blurry since it's not always clear whether something can't be done because you've reached a fundamental limitation vs. having a clever enough design.

It's possible someone will eventually solve the problems in these areas (assuming there aren't fundamental reasons we're just not aware of yet why it's impossible) but IMO the odds are low and you need some sort of breakthrough.

> there are also no real signs that a quantum computer will beat a "classical" computer at anything any time soon

John Martinis' group is planning to do that within the next two years [1][2]. Ten years might be realistic for solving useful problems faster and cheaper, but the esoteric speedups are right around the corner.

1: https://arxiv.org/abs/1608.00263 2: https://www.youtube.com/watch?v=kgMWommXxU8

Just watched the talk. Definitely an interesting direction.

One question I'd have is whether quantum simulation is indeed the fastest way to solve the quantum speckle problem on a classical computer. The other related thing to think about is whether the output of a quantum speckle computer can only be explained by the large state space and quantum mechanics or whether there are alternative explanations.

So definitely an interesting direction but even if they're successful we're still not quite there. When we have a quantum computer doing useful work with exponential speedup over classical computers that would be something and even then we only have a very limited set of algorithms at the moment.

> whether quantum simulation is indeed the fastest way to solve the quantum speckle problem on a classical computer.

They give arguments for this in the paper. Intuitively, if there's some fast way to classically sample a random quantum circuit then that means most quantum circuits should be easy to simulate classically. But that doesn't seem to be the case: we don't have algorithms that go fast on most circuits but go slow in highly specialized cases. It's the other way around.

> whether the output of a quantum speckle computer can only be explained by the large state space and quantum mechanics or whether there are alternative explanations.

Given that all the engineering was based on quantum mechanics, it would be very surprising if the machine worked based on something else. It's quite difficult to make things accidentally work like that. It'd be like trying to build a car, confirming it can take you from place to place, then realizing that actually you made a hang-glider by accident.

> With quantum computing there are also no real signs that a quantum computer will beat a "classical" computer at anything any time soon. It always was and still is very difficult to get quantum systems to scale and you need a large enough system to make a difference and it may be impossible to do so. Again a lot of this is more of an engineering problem though admittedly the line between physics and engineering can be blurry since it's not always clear whether something can't be done because you've reached a fundamental limitation vs. having a clever enough design.

The mentioning of quantum computers is very interesting in the following sense: If one were able to build a sufficiently large quantum computer, this would provide strong evidence that the (at the moment rather hypothetical) theory whether quantum mechanics is rather an emergent phenomen by a deterministic process (cellular automaton)

> https://arxiv.org/abs/1405.1548

which the Nobel laureate Gerard 't Hooft worked on for the last years is probably wrong. To quote p. 79-80:

"Such scaled classical computers can of course not be built, so that this quantum computer will still be allowed to perform computational miracles, but factoring a number with millions of digits into its prime factors will not be possible – unless fundamentally improved classical algorithms turn out to exist. If engineers ever succeed in making such quantum computers, it seems to me that the CAT is falsified; no classical theory can explain quantum mechanics."

On the other hand, if we seriously get into trouble building a sufficiently large quantum computer (despite our best efforts), this would at least to me provide evidence that 't Hooft is on something - since that this is a prediction that his Cellular Automaton Interpretation of Quantum Mechanics provides.

I think that the largest simulation of a quantum computer on a super computer is about 40qbits. Therefore quantum supremacy (weak...) is 41 stable qbits. There are numbers of groups out there that are building (working on projects with the funding to complete) machines with 100's of stable qbits now, so my argument is that quantum supremacy of the weak sort is a matter of engineering. I think that quantum computers that factor primes with 100's of digits (like 2048 bit ones) requires 1000's of qbits, say a couple of orders of magnitudes beyond the machines that are currently in train to be built.

The Hooft thing is that there are limits on computation due to the process of the universe and that if a classical computer that was at the plank scale, runs at plankian speed (help help I am using terms I can't understand!!!) and covered the whole of the universe couldn't do it from the beginning of time to heat death... then nor can an arbitrary QC. Shores algorithm does factors in (log n)^3 maybe we'll find that there are physical limits on the scale of QC which hold it below hundreds of millions of qbits where a plankian computer could be synthesised, and/or it may be that such a computer can't run at high clock speeds, and because of this BQP algorithms can't be run on problems that are not allowed in the sense that they somehow solve the universe.

I don't think anyone has any reason to believe this is true in an engineering sense but there is every reason to believe that large QC will be very, very difficult in an engineering sense and I think that the people building one do expect to run into problems of this sort eventually, but at the moment they are delighted to find that they are able to do things at the lower end that will create devices that are going to be transformational for reasoning about some components of the physical universe - for example simulation of the interactions of very large molecules - maybe mapping seconds of a virus interacting with a cell membrane using months of QC ?

One of the things that the fusion community is trying to do is scale down the size of machine that is required, basically the view is that the volume of facility determines the cost (interestingly the stats presented seem to support this) and by building compact devices the price of the machine will decline accordingly. This will enable commercial implementations - if it works!

One thing to remember with quantum computation is that "large enough to be superior" is measured in 10s of units of quantum resources. That doesn't seem so far away.

Not sure if you're serious or joking (the parenthesis suggests to me serious), but AFAIK it's a well known saying that fusion power is in permanent state of "to be delivered commercially in ~15 years" since '50s or something like that...

Edit: ah, yes, and also the slightly more generic mandatory XKCD: https://xkcd.com/678/

I think his point was that it's now 15 years away, which is not the 20 years away it was 5 (and 20) years ago. In that sense it is significantly closer. Many of the real challenges may remain monetary as solar becomes cheaper, and funding of fusion has never been particularly high (perhaps $10-15B over 50 years). However, it is ultimately a complementary technology (as are Thorium reactors and others) for scales, places, and times when Solar or Wind aren't appropriate.

http://www.huffingtonpost.com/2015/01/20/fusion-energy-react...

The real surprise here might come from Lockheed-Martin's fusion project.[1] They're pursuing an unusual approach to magnetic containment. It's not a tokomak, and it's not a stellarator.

Many people have treated this as a hoax, or a joke. But the work is being done by Lockheed-Martin's Skunk Works, which has quite a track record. (The U-2, the SR-71, and stealth aircraft came out of there. ) Lockheed-Martin is doing this with their own money, and last year, when asked about it, the CEO said it was doing well enough they were putting more money in.

[1] http://www.lockheedmartin.com/us/products/compact-fusion.htm...

> deliver it commercially in ~15 years

According to the researcher translation table, fusion is still a long ways off. https://xkcd.com/678/