back to article Tiny quantum computer plugs into top Euro supercomputer – because why not?

What happens when you plug what's said to be a quantum computer into Europe’s most powerful supercomputer? That’s just what researchers at Finland's government-sponsored technical research institute VTT aims to find out. On Tuesday, VTT announced it has connected Helmi — which translates to Pearl and is a basic quantum …

  1. Anonymous Coward
    Anonymous Coward

    Not quite getting the point of this...

    5 qubits isn't enough to do anything meaningful (and to get a scale of this IBM bought out a 5 qubit chip back in 2016) so I don't quite get why you'd want to add something a bit toy to a real supercomputer. There's an argument that this is useful exploratory work but in practice how one would use a hybrid computer at scale would evolve in actual usage solving actual problems.

    Its interesting but comes over as more headline than useful in practice.

    1. cosmodrome

      Re: Not quite getting the point of this...

      Well, they have turned a boring "one of the most" powerful supercomputers into the single most powerful experimental toy for a hand full of eggheads. Successul upcycle. Progress!

    2. Anonymous Coward
      Megaphone

      Re: Not quite getting the point of this...

      Often supercomputers, at least those not run by defense or security related agencies, are used by boffins to do, as you put it, exploratory work.

      Pure science is not about actual usage solving actual problems - that's beancounter thinking.

      I'd normally just give the boffins a pint but I've elevated this to a megaphone.

      1. Anonymous Coward
        Anonymous Coward

        Re: Not quite getting the point of this...

        You've muddled up two separate things. One is the use of the Lumi supercomputer (as a fast bit of compute infrastructure) to solve problems, or as you put it, exploratory work.

        The other (which is what I meant) is exploratory work in the capabilities of that infrastructure by adding a 5 Qubit quantum device, which everyone acknowledges is absolutely useless in terms of doing actual work. What they have done is made their supercomputer capable of supporting a 5 Qubit device by putting in drivers, as in: 'all the drivers of the future of computing seamlessly integrated and ready to be utilized.' We heard this when supercomputers moved to hybrid GPU architectures.

        Adding an extra flag to your job queuing mechanism and bolting a 5QBit chip is not 'boffin' material. Its a publicity stunt.

        There are no meaningful quantum computers ready to be attached. And if there were (lets say the stability problem is solved and we can have 100 MQubit computers) I strongly suspect that the architecture wouldn't support the hybrid compute effectively. Supercomputers are generally highly optimised and tightly coupled from a design perspective, and you don't just bolt stuff into a tightly coupled architecture, whatever that is.

  2. Lars Silver badge
    Happy

    About LUMI

    LUMI is co-funded by the EuroHPC Joint Undertaking and the LUMI Consortium, which is composed of the following countries: Finland, Belgium, Czech Republic, Denmark, Estonia, Iceland, Norway, Poland, Sweden, and Switzerland. The total budget is €144.5 million.

    https://en.wikipedia.org/wiki/LUMI

  3. Anonymous Coward
    Anonymous Coward

    Stating the obvious

    A photon hits an electron, what properties are you detecting? a) the properties of the photon. or b) the effect the photon had on the electron?

    Well b) obviously. If the photon had no effect on the electron, then it would not be detected at all, so it must be b).

    So, those effects (things like wavelength, polarization, all of them), are not the properties of the photons, they are the net effect the photon had on the detecting electron.

    Put simply, a photon is not red or blue, it interacts with an electron, that *interaction* is red light or blue light depending on the motion of the electron. It's wavelength is just the net effect of one on the other. The apparent wavelength is just a net oscillatory motion between the electron and the photon, all the other properties are also oscillatory motions, things like polarization is a net oscillation effect. Spin is just two such oscillations that are not-co-planer.

    So of course you cannot determine the net effect that photon will have on an electron till you select the electron you will measure it against. Your photon is interacting with multiple electrons around it all the time, and their motion means the photon appears to have different properties on each. You choose the observer you measure against, and can determine the net effect of the photon on that chosen observer.

    You are not setting the properties of that photon by measuring it, you are simply selecting your reference electron (in this example its an electron, although its actually a point in an oscillating field).

    The photon is definable at macro scale, those quantum scale oscillations must add up to that macro scale motion. So it is fully defined.

    Electrons move predictably at macro scale, all those little oscillations add up to its motion, so it is fully defined. Lots of those oscillations cancel out, hence it is not moving.

    **********

    Which implies electrons have an oscillatory motion (the difference between our red-shift electron and blue-shift photon is an oscillation, which comes from the electron, so the electron is following an oscillating pattern).

    Which implies electric is an oscillating force (interaction with the electron is via its field).

    Which implies the underlying non-oscillating force propagates infinitely fast (all direct and indirect paths from our electron must reach the same point at the same time). i.e. H0 propagates at infinity, there is no information paradox, and black holes definitely do interact with the space around them.

    Which implies time is relative to the observer (think of time as an oscillation in an atomic clock, that oscillation must also be a net effect relative to an observer, so time must also be relative to the observer's motion). i.e. Einstein and Shrodinger were looking at the same effect.

    Which implies our electron must be in a repeating pattern, it is 'stationary' relative to the photon, so overall it moves left as much as right, up as much as down.

    Which implies speed of light is relative to an observer (another net motion, must be relative to the observer).

    Which implies the event horizon of the black hole depends on the observer, move faster and it moves away from you.

    Loads of fun there, e.g. A B C are 3 stars falling into a universe-eating black hole. 'A' thinks the event horizon is between A & B, B see its between B & C. If we are B, then all paths lead to the B/C event horizon, so the blackness around the universe is the *inner* event horizon. Conversely *no* paths lead to the outer event horizon (A/B), so we cannot even perceive it. There is no difference between the inside and outside of a black hole.

    We can go on and on and on. But first accept the simple thing I am asking here: What you think of a photon is actually its effect on an observer. What you think of as an electron is its effect on other particles. None of the magical 'dead-alive-cat' is real. Get past it.

    1. sw guy

      Re: Stating the obvious

      What is obvious after reading the post is that your sense of logic is broken

  4. vtcodger Silver badge

    60 years since ...

    It's been 60 years (59 if you're picky) since Richard Feynman told us that nobody understands quantum mechanics. Still true I think.

    I certainly don't understand quantum, but one of the few things I might understand about quantum computing is:

    -- About the only thing quantum computing can (currently) (possibly) (on really good days) do is superfast Fourier Transforms. And the only thing those SFFTs might be good for is cracking some forms of cryptography.

    Can anyone who actually understands this stuff straighten me out? And, are only 5 bits -- even 5 super duper quantum bits -- likely to be good for much of anything?

    1. martinusher Silver badge

      Re: 60 years since ...

      >And, are only 5 bits -- even 5 super duper quantum bits -- likely to be good for much of anything?

      Learning how to use 500 qbits, possibly?

      The article is a bit vague about what they mean by 'connected to'. This gives it more of a PR feel than a practical purpose -- after all, assuming the supercomputer's purpose is supercomputing and not just chewing through massive amounts of data it doesn't really matter much where the user and peripherals sit. Its probably just convenient to have 'people who understand how to fiddle with qubits' and 'people who know how to write software for / configure / use a massive computing array' in the same location.

  5. vtcodger Silver badge

    60 years since ...

    It's been 60 years (59 if you're picky) since Richard Feynman told us that nobody understands quantum mechanics. Still true I think.

    I certainly don't understand quantum mechanics. And I'm almost as hazy on Quantum Computing, but one of the few things I might know about quantum computing is that about the only thing quantum computing can (currently) (possibly) (on really good days) do is superfast Fourier Transforms. And about the only thing those SFFTs might be good for is cracking some forms of cryptography.

    Can anyone who actually understands this stuff straighten me out? And, even if Quantum Computing can actually do wonderful stuff, are only 5 bits -- even 5 super duper quantum bits -- likely to be good for much of anything?

  6. John Smith 19 Gold badge
    Coat

    Still. Serious Floppage from a (relatively) small budget. As motor racing, so IT. If you want to win

    Hire some Finns

    Thank you and goddnight

  7. John Smith 19 Gold badge
    Unhappy

    Actually there are reports that as little as 4 qubits could be useful...

    Something called the Nature in 2019 It's about using a conventional/quantum hybrid to solve the Harrow-Hassidim-Lloyd algorith, which apparently is a way of solving linear equations, of the sort that show up a lot in CFD simulations.

    The question of course is how big a problem can a 4 QB machine solve (or rather a HPC array with addon) and if it's quite small can the the big arrays be split in a way that allows the addon to handle them :-(

    If the anwer is "Usefully big" and/or "yes" then they are in on the ground floor of something quite big.

    But I've seen so much Quantum-Computers-will-change-the-world BS that I'm pretty doubtful.

    Time will tell.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like