http://mpcollab.org/MPbeta1/node/165
I doubt we'll stay at the level of hundreds or thousands of qbits / system for long; I think millions is feasible and will be done in short order. I doubt that it'll take 20-30 years either, more like 9-15 for it to be pervasive in the marketplace. The materials science nanotechnology is just about ready and commercialization of it won't take long after the R&D is done to a workable level.
Think about a modern GPU like the ATI 4850, it has 800 ALU units in there, each one having something like fifteen or more thirty two bit wide registers and other storage associated with it totalling several hundreds of bits per ALU, and perhaps half of a million bits per chip just for the ALU inputs and outputs.
So having millions of qbits active on a chip would be in some ways needed to have the memory storage capacity needed to work on problems of a certain data input and output size even if there were relatively fewer quantum 'ALUs' than qbits of storage feeding and connecting them.
Parallelism will be important; if you can network a few hundred QBITS into a little quantum microcomputer / QALU then you can just put lots of those modules in parallel to address many bigger problems or to have higher performance.
The way the wavefunction collapses during measurement is just a reflection of Heisenberg's principle of uncertainty; observing one independent state attribute of the system increases the uncertainty of the other independent state attributes of the system.
Think of a planet orbiting in a circle / ellipse not spinning on its axis. It can have an up/down component of its motion and a left right component of its motion. Just like at 3 and 9 o'clock the clock hands are moving purely up/down, and at 12 and 6 o'clock they're moving purely left/right.
If you want to test whether it is moving left to right you can take a wall that is oriented up/down and place it in its path. If it crashes into that wall, it must've been moving left to right. If you want to detect up/down motion you can take a barrier aligned left/right and place it in its path. If it hits that, it must've been moving left/right. After you collide it, though, you can't tell by a future measurement whether it was moving in the other axis because you've already disrupted its path completely by measuring one of its linearly independent motion components.
The overlapped quantum states are linearly independent states of the system so they don't interact with each other, but by precisely measuring one of those attributes you're disrupting the others. You can see if a photon is polarized up/down with an up/down linear polarizer. Photons that aren't polarized up/down don't get through, so you can't repeat the experiment with a left-right polarizer because the photon didn't get through the first up-down polarizer to begin with.
Originally posted by: firewolfsm
Recently I've been more and more interested in quantum computers, read some articles on caltech and wiki and now I'm curious to see what any of you think or know about them.
I still don't really get how a superposition of n states collapses to the correct calculation after a measurement. But everything else seems to make sense.
I imagine a computer in 20 or 30 years with a quantum computer next to a digital processor in the same box.
Also, I realize we're basically up to the vacuum tube computer stage of quantum computer development, but is there of a possibility of having millions of qubits in one system eventually? Thousands are probably more than enough for any application though.
Facebook
Twitter