back to article Google demos image rec 'quantum computer'

Google says it has developed a kind of quantum computer capable of identifying objects that appear in digital photos and videos. According to the company, the system outperforms the classical algorithms running across its current network of worldwide data centers. Hartmut Neven, Google technical lead manager for image …


This topic is closed for new posts.
  1. Anonymous Coward

    Let it be known to all:

    I hereby copyright all images of me that I have not signed a release form. I do NOT give Google permission to use, distribute, redistribute or share with any of their partners in crime. So, if I appear in any videos, satellite images, or photographs that I have NOT authorized Google to use, the damn it, keep your mitts off!

  2. Charles Calthrop
    Thumb Down

    you lost me at

    "They're stored in a quantum system, such as, say, the magnetic spin of an atomic nucleus. An "up" spin might indicate a 1, a "down" spin a 0."

    But can the fucker drink 8 pints of bitter like I just have

  3. Shady

    But can it...

    ...recognize privacy concerns? Or this emoticon? ..|..

  4. Anonymous Coward

    @ Charles Calthrop

    The question you pose goes to the very heart of quantum theory, and was formalised in a backstreet Copenhagen drinking den late one night in 1927. It is, of course, more commonly known as Schrödinger's Pint, the primary variant of which is still ardently debated in countless hostelries the world over.

  5. Tony Paulazzo

    3 points

    more free porn better tailored to taste

    wouldn't the qubit be in both a 1 and 0 state until someone looked at it?

    google would use this with the desire to do good, but the power of the ring is... sorry, wrong article

    when this quantum pc achieves sentience via the power of the web will it destroy all mac type products and their owners? only allowing linux to survive cause it finds it amusing

    wouldn't the heisenberg uncertainty principle mean it's wrong 50% of the time?

    I'm really bad at math

  6. DaftWullie

    That reminds me...

    of a quote I once found by my PhD supervisor, Artur Ekert,

    "I do hope not to live long enough to see the launch of Quantum Windows"

  7. Anonymous Coward


    The title would lead one to think that some image recognition was demonstrated. Was this part of the article accidentally left out?

  8. Mike VandeVelde

    "increase our appreciation for the structure of reality"

    Yes, hmmm.

    I say: what better way to fund this madness than by milking marketers? Milk them, milk them hard!

    If cool shit comes out the end, and all I have to do is ignore ads to get there, then bravo!

  9. Lance 3


    Windows already is, it is an OS that doesn't operate.

  10. jubtastic1


    I thought the whole point of a bits being fixed in one of two possible values was that you could write some data to them and be pretty sure it would be the same when you read it back in later.

    Could someone explain (in numpty terms please), how 'all states at once' is a useful thing?

    also, where is the *whoosh* icon?

    1. Adam 41

      "Quantum" Image Processing

      Its because a quantum computer can effectively look through all possible values at once, because the bits its using to search can take all values. Basically it can search a database of size n in much less than n queries.

      Having read the NIPS demo paper that Google reference I have a few points to make. First they don't actually use the quantum computer to do image recognition, they use the quantum computer to devise the rule set which is then used by normal computers to search images. Second, their main results come from the simulation of the quantum system on normal computers rather than running it on D-Wave's hardware (as that hasn't got enough functioning qubits yet). Third, there doesn't seem to be all that much computing going on in the quantum computer, as they just initialise it to a flat state and then slowly poke it until its working on a similar problem to the one they want to solve, at which point they let it decohere and read off what the result is.

      Still, its damn cool.

    2. Greg 10


      Yes, one can easily explain.

      "All states at once" mean it's a% state A, b% state B, c% state C and so on.

      With a, b, and c of your choosing, if all goes well.

      I think you can see quite easily how it could prove superior to have the choice of either A, B or C Well, actually it's obviously much more complex, but the point is still that it's not just "it's everything at once".

      It's like saying when someone says "you know, before, we had Red and Yellow dots, but now we can have both at once on the same dot". It may seem absurd at first, but then you realize it means you get access to all shades of Orange, which will clearly hold more information (as for computing algorithm, I'm completely unable to explain, but you get the idea anyway)

  11. Anonymous Coward

    Search perfection guaranteed

    Heh, so now Goog will start claiming all their search / image results were 100% correct when their quantum computer calculated them, therefore if the results are wrong its because the USER LOOKED at them ;-)

    And G if you really have a quantum computer, can't you use it for something better than computing pron, i dunno, cure for cancer, world poverty etc ... just a suggestion.

    1. Anonymous Coward
      Anonymous Coward

      I can just see quantum computing being used for

      cracking encryption keys (for dnssec and the worlds banks) for more effective phishing attacks

  12. Anonymous Coward

    Mere Mortals

    I'm afraid this appears to be the point where mere mortals get left behind. Quantum physics is at the edge of reality by its very nature so using such things for computing could go end up taking things in all sorts of directions all at the same time.

    Wonder when they'll crack uncrackable encryption, if they're doing things that can't be done with unlimited 'classical' computers its only a matter of time. Hmmmmmm......

    1. Michael H.F. Wilkinson Silver badge

      Uncrackable: no, fiendishly difficult: yes

      A one-time pad code cannot be cracked at all, but anything relying on prime decomposition can be cracked by quantum computers in quadratic time (in the number of bits) as opposed to exponential time.

      On the up-side, the very act of intercepting a quantum encrypted message collapses the state of the message, so interception is immediately obvious.

  13. Anonymous Coward
    Big Brother

    A working quantum computer in the hands of a civil company?

    So the NSA most likely has a similar chip for cracking RSA...

    1. copsewood

      Very likely indeed

      Though how far they have got in connection with the use of Shor's algorithm in cracking the RSA problem is an interesting and open question. I'm not a mathematician, but my quick scan of the following article suggests a 2048 Qbit computer is needed to crack a 1024 bit RSA key:

    2. Ray0x6

      I dunno

      Part of me thinks that in a massive subterranean cavern in Nevada are a bunch of the world's best scientists and engineers that you've never heard of, constructing these devices for the NSA out of our sight. But another part of me thinks that If the rest of the world put together can't figure out how to build a quantum computer.

  14. ChrisInAStrangeLand

    Problems down here... no wait, that's NASA.

    I wasn't aware poverty was an NP complete problem. My understanding is that poverty is caused by angry sounding politicians trying to 'help' their people by bringing the successful land owners and businessmen to justice.

  15. Anonymous Coward

    Turing says...

    Is this any more powerful than a classical Turing machine, or does it just compute the same things faster? If the qubit superposition allows the construction of an actual nondeterministic Turing machine, it has dramatic consequences, such as P = NP for practical purposes.

  16. Mike Kamermans


    In a nutshell: quantum computing algorithms work with different operations from classical systems. In classical systems you can set a bit to either 1 or 0, and then you can perform gate logic with them. In quantum computing algorithms, you can do most gate operations except copy patterns (because all quantum gate operations must be reversible for reasons that have to do with a need for humans actually being able to understand the math that underpins the algorithms). In addition to the regular gate operations, there's also a set of operations that can "shift the balance" between the 0 and 1 state of a qubit so that it's, say, 50% 0 and 50% 1. then any further gate operation will actually operate on both those states. negating the qubit will make it 50% 1 and 50% 0 for instance. As long as you don't read the qubit befor your algorithm is done, this is great: it means you can apply one operation to a series of qubits and -if your algorithm is welldesigned, and extremeffs they are not easy to design- get out a qubit sequence that contains every possible correct answer, and none of the wrong answers.

    The problem is you can only 'select' on of these answers, and you can't pick which one. once you measure the qubits, you get only one of the numbers, so as long as you only want "a right answer" rather than "a specific right answer" (for instance, factoring a number into its prime constituents) , or you designed your algorithm in such a way that the final qubit pattern is the only possible answer, job's done.

    The real problem is that qubits are unreliable. Unlike classical bits, they tend to wobble about a bit and not always contain what you want them to contain, so quantum algorithms operate on the principle of "it's bloody unlikely that if I run it ten times and it gives the same answer ten times, that answer is wrong". Unlike a classical algorithm, you have to run a quantum algorithm quite a number of times before you can say that you're 99.999% confident the answer is correct. However, since the time required to run the algorithm can be an order of magnitude less than the classical approach, that's actually perfectly feasible.

    If we had an actual quantum computing. Because so far, we don't. We just have books on the theory and math behind it, that shows that it will kick ass if we could ever build them, but the inability to copy (and thus duplicate) any bit pattern makes quantum algorithms hilarious complex. Hell, the most famous quantum algorithm (shor's algorithm) seems to have just been discovered mostly due to a stroke of luck (or, in engineering analogy, randomly hitting the keys and discovering you wrote a webbrowser)

  17. Rocketman

    The Multidimensional Universe Angle

    Let's suppose that each of the possible values of each of the qubits and each possible combination can be thought of as a possible universe and let's suppose that the universe works like that and how a quantum computer can work is that two such quantum computers work together as a kind of differential comparator thus allowing the comparison to occur in 'quantum' space and the trick would be to achieve a comparison value that would have a rather high probability of being true so that while destroying the quantum states, the comparison was done in such a way that an answer of high probability would result.

    It seems to me that this quantum computing could lead to an understanding of the multidimensionality of the universe, which I think is immensely cool. Because if this quantum stuff is real, isn't that what we are looking at?

  18. blackworx



    That is all

  19. Neil Stansbury
    Thumb Up

    Video clips

    This is of the training process:

    And then presenting their findings:

  20. Gordon Pryra

    Cracking Keys?

    Is that the best people can come up with?

    If it can crack current encryption, then just use it to create a new key that its peers can't crack

    Whats more scary is the fact that if this thing is being tested with systems its designed to learn from, then it Skynet time.

    Add to this, the fact that its plugged into the heart of Google and all they control....

    When Google start ordering heavy construction gear, its already too late to head to the hills.

  21. iamapizza

    It won't work

    I don't think their quantum searches will work - they'll change the results by observing it!

  22. ThomasF
    Big Brother

    A Quantum Leap for Google Skynet

    It's John Connor time again it seems as "Rage Against The Machine" takes over the Google Xmas Charts

    Google seem to be pushing the envelope everywhere. All them guys in Montana fighting against the UN will have to move on as The Google men look like they will rule the Qubit World.

    If there is a world after Cern and the Mayan's are finished with it.

    "Quantum Adiabatic Algorithms" Nuff said:+~

  23. Matt 21


    What's your name George Bush?

    When the dollar was higher it would have been much cheaper to pay a load of Indians to surf the web and identify the images for you. Of course now it's cheap as chips to get a load of Yanks to do it, if only they could be trained to the right level :-)

  24. The Indomitable Gall

    Classical and quantum interactions...

    "The rub is that if a quantum system interacts with the classical world..."

    If quantum computation fails based on the classical world, surely we should stop calling it "quantum computation"? That blend of both classical languages surely means that even talking about it destroys the effect!!!!

  25. John Bishop

    Quantum encryption/decryption

    Quantum computing seems to allow both the generation of an uncrackable code and a computer that can crack it

    My head hurts...

  26. Destroy All Monsters Silver badge

    HELL NO!

    "If one takes a closer look at such capabilities one realizes that they often require the solution of what mathematicians call hard combinatorial optimization problems. It turns out that solving the hardest of such problems requires server farms so large that they can never be built. A new type of machine, a so-called quantum computer, can help here."

    Okaaayyy... if these are the NP-hard problems, it is very unlikely that the a Quantum Algorithm will help. Very, very unlikely.

    See also:

  27. Stephen Hunt


    Now if they could only train it to identify vapourware...

This topic is closed for new posts.

Other stories you might like