New Idea Could Lead to Quantum RAM

KentuckyFC writes to tell us that scientists in Italy and the US have designed a new method of retrieving information from quantum memory that could allow them to create "Quantum RAM". "Giovannetti's idea is to send the address down the branching tree of connections in such a way that it only affects one switch at a time. The first address qubit sets a switch at the first branching point to go one way or the other; the second qubit is sent that way and sets the switch at the next branching point, and so on. The total number of entangled quantum systems is smaller, and they are not so susceptible to interference, allowing information to be retrieved from memory intact."

Comments

talk about density

[downix]

atomic-scale memory would create huge waves.

It also could help out on the heat issues as well.

I mean, think about how many atoms are in a normal piece of memory.... yeouch that's a lot of RAM!

binary trees

[ch0ad]

could this be used to implement extremely efficient binary trees? the structure sounds ideal to be but im hardly an expert.

Quantum address?

[aadvancedGIR]

I know I don't really^H^H^H^H^H understand quantum computer, but don't their output rely on some random effects to quickly generate statistics that would take ages with regular brute force computing? So would a quantum RAM provide the stored value only X% of the time?

Re:Quantum address?

[Anonymous Coward]

A quantum register can be in a superposition of multiple states. For every possible state, there is a corresponding number called a probability amplitude. The square of the absolute value of the probability amplitude is equal to the probability that when observed, that state is the one you see, so naturally, the squares of the absolute values have to add up to 1. A quantum gate maps a single state onto a superposition of states.

Maybe an example will make this clearer. A quantum state is often written as a ket: |0> and |1>, for example. The Hadamard gate maps states like this:

|0> to (|0> + |1>)/sqrt(2)
|1> to (|0> - |1>)/sqrt(2)

If you feed |0> to a Hadamard gate, then the result will be a superposition of 0 and 1. 0's probability amplitude will be 1/sqrt(2), as will 1's. Feeding |1> into it will give you the same thing, except now 1's probability amplitude is -1/sqrt(2). (A probability amplitude can be a complex number.)

So suppose, now, that we feed (-|0> + |1>)/sqrt(2) (that is, 0 with amplitude -1/sqrt(2) and 1 with amplitude 1/sqrt(2)) into the Hadamard gate. You can just substitute the result in for the kets, like this:

(-|0> + |1>)/sqrt(2)
(-(|0> + |1>)/sqrt(2) + (|0> - |1>)/sqrt(2))/sqrt(2)
(-(|0> + |1>) + (|0> - |1>))/2
(-|0> - |1> + |0> - |1>)/2
(-2|1>)/2
-|1>

The result will be 1 with a probability amplitude of -1, corresponding to a probability of 1.

So in practice, each qubit will probably be a single particle or something. With 8 qubits, you have 2^8 probability amplitudes that a classical computer would have to keep track of separately. Unfortunately, you can't go around doing just anything you want to these probability amplitudes, like manipulating them one by one, but manipulating them in certain ways can land you with a nice, fast algorithm, like Shor's algorithm, which can factor integers (which is useful for breaking certain types of codes) faster than any known classical algorithm. How does it work? I have no idea :-)

Re:YAUQA: Yet another uninformed Quantum Article

[kebes]

a quantum gate or quantum computer is only capable of probabilistic answers

I don't think that's true. Yes, many quantum processes are fundamentally probabilistic, but that just means we need to avoid those processes when building quantum computers. The intended design for a quantum computer is to use unitary (invertible, deterministic, etc.) operations for the quantum gates.

The main roadblock to keeping the gates unitary (i.e. keep the error rate low) is to have the switching occur faster than the decoherence time (the timescale over which the delicate superposition decoheres into a random probabilistic mixture). This is certainly a difficult issue to solve, but in principle it is possible. The small-scale quantum computers that have been built to date were able to solve small problems deterministically.

As a practical point, it may turn out to be very difficult to build a quantum computer... but as far as I know the intended designs of quantum computers are not to yield probabilistic answers and then to average them, but to maintain coherence long enough that the final answer is deterministic, with an acceptably small error rate.

YAUQC: Yet another uninformed Quantum Comment

[Anonymous Coward]

In particular, a quantum gate or quantum computer is only capable of probabilistic answers. That is, each gate only has a slight predisposition to give the right answer. How you'd use unreliable gates to do say a 32-bit address decode is a bit of a brain-teaser. Without huge amounts of error-detection and correction, there's only a 1 in 2^32 chance it will access the right memory cell. We need like 1 error in 10^10, a 10^19 shortfall in reliability. I'm loath to slough off nineteen orders of magnitude.

The fundamentally probabilistic aspect in quantum mechanics is only in measurement (and even there, there are situations where measurement results are completely non-probabilistic; especially it's completely possible to emulate a classical computer on a quantum computer). For typical quantum algorithms, the only time you do a measurement, you do it at the very end. Especially this quantum RAM proposal does not in any way contain any measurement, therefore on an ideal quantum computer, it works completely deterministic. Especially, if your qubit pattern really adresses only one memory cell, it will be reliably fetched. But in general, your qubits will address a superposition of memory cells, and will therefore cause reading a superposition of values (there's still no randomness in here; only if you were to measure that stuff, you'd get randomness in; but even then, in the general case it's not a 50% chance to get the bit right; if that would the case, not only quantum computing, but any meaningful prediction in the quantum world would be completely impossible even from the start).

Now, in the real world, there are random influences besides the fundamental randomness in measurement (simply because you cannot completely isolate the system from the environment), and those influences are indeed a big problem of quantum computing. Now that's exactly the point of what the authors did: From what they claim, their scheme is more robust against those influences because there are less gates interacting with the qubits, i.e. less ways for the environment to mess with the system.

BTW, they also claim their memory adressing scheme could give energy savings on classical computers (since classical computers provably can be built, their work could be useful even if there will never be a single quantum computer of more than a few qubits).

Re:Talk about it if you must, but DO something

[Jeremiah Cornelius]

Were you a g*ddammned hall monitor or something in a past life?

What are you talking about? I replied to a non-sequitur about doing "something".

I did "soemthing".

I posted.

About a funny idea that occurred from mis-reading the post. Quantum DRM is a possible misapplication of the crypto possibilities in quantum computing. But the existence of quantum RAM suggests that the problem might be moved into an area that was not previously accounted for in the speculation.

I was appealing for folks with insight to explore this possibility, or to demonstrate where this was tangential or irrelevant to core cryptography problems.

That is not nonsensical.

Oh, and I think you might need to get laid, or something.