Slashdot Mirror
A Pair Of Quantum Computing Articles
Will G writes: "3DRage has posted an article entitled "Quantum Computers: How they work and How they will effect us" by Alan Cline. Not only can quantum computers run one billion times faster than typical silicon-based computers, but also theoretically, they can run and consume no energy. That being true, quantum computers could obsolete the silicon chip much as the transistor did the vacuum tube. This paper is intended for the general reader, and explains basic quantum computer features, and the paradoxical effects quantum theory produces in a practical world. This paper discusses how quantum computers originated, the inevitability of their use, and how they differ from classical computers." An interesting nugget to add to this comes from leelaw2000, who writes: "New Scientist have published this little news story about the development of a kind of quantum shielding that might help the development of real quantum computers. Now if they can just get Quake on it ..."
I understand it differently. AFAIK, Landauer showed that enmergy must be spent to erase information, not to clear a bit. You are erasing information if you use a gate with more input wires than it has output wires. For example, a clasical NOT gate has one input, one output, and no info is erased because you can always tell what input caused a given output; it's reversible. But a classical AND gate could cause a 0 output in three possible ways (00, 01, 10), and you can't tell which; it's irreversible, because you've erased information. Read http://www.qubit.org/intros/compSteane/qcintro.htm l for some more info.
Also, someone should note that the energy savings from reversible computation are real but very, very tiny. Chips would have to get 1 million times more efficient than they are now for the energy costs of (current) irreversibility to manifest themselves. And if you expect a quantum computer to operate without tons of expensive, high-powered supporting equipment around it (NMR machines, optical pumps, liquid helium-cooled ion traps), you'd better add a couple more decades onto your time estimate.
Zero-energy computation isn't anything new, in theory; we've known what must be done to achieve zero-energy computation for a long time. We just haven't quite been able to figure out how to do it.
:)
In principle, setting a bit requires no expenditure of energy; it's clearing the bits that requires an energy expenditure. So, provided you can figure out a memory design which permits that bits be set and never cleared, you can achieve zero-energy computation.
Note that I'm using "requires" in a very narrow context here. Setting a bit requires no expenditure of energy, but all the computers we have right now expend energy to set bits. That's a limitation of design, not any thermodynamic limitation we're currently aware of.
All of this comes to you courtesy of some long-ago college courses on the physics of computation. I may be misremembering quite a bit.
In real life, we don't have systems accurate enough to deliver one photon to one atom (or nucleus). Instead, we play the odds and bombard the q-bits with a very large number of photons until it is in the proper state. All the other photons are lost.
Technically, they could capture all the photons emitted by the q-bits and return them into the system at a later time. But I don't think that will happening any time in my lifetime!
-----------------
In other words, if there is no thermal dephasing, you can operate with no energy consumption so long as you never look at the output, but there is a rigorous minimum value of energy that it costs to look at the output. This limit is set by basic thermodynamics and is inescapable.
In practical terms, cooling the computer to feasible cryogenic temperatures will consume lots of energy even when the qbits do not. Moreover, the fact that you will run the computer at finite temperature makes it necessary to apply error-correcting codes to compensate for thermal dephasing. Error-correcting steps are irreversible and thus consume energy during the calculations.
what I was always curious about was whether quantum computing could be put into non cpu environments, still processing, but not as general purpose. For example, could a quantum chip be put onto a 3d card, and make it work a million times faster? Or could it be put to use in network switches with regard to 100% optical switches pushing us into 1TBit networking?
---
Video meliora proboque deteriora sequor - Ovidius
for Quantum Computing Purposes...
Skip this if you've had even Physics 101.
First of all, Quantum Theory as we know it has been devised over the last century. I could name a lot of famous scientists names like Heisenberg, Schroedinger, and Fermi, but you don't care so I won't.
The meat and potatoes of quantum theory is this: All particles, no matter what the size, act as both a wave and a particle. According to research, either the location *or* the mass of a particle may be known at any time.
Also, as we all know, wave interfere with eachother. If the crests of two waves overlap, they grow. This is referred to as 'Constructive' interference. 'Destructive' interference happens when a crest of one wave overlaps the trough of another wave. This gives rise to many observable phenomena, such as diffraction lines you can see when you stare at a bright light through your eyelashes. This is what causes 'ice rings' around bright lights in cold weather and the occasional 'moon ring'. It's also why you have to have your surround sound speakers positioned just so, so that they don't interfere with eachother.
Early experiments where researchers shot electrons through tiny holes in a lead sheild and onto film created similiar diffraction patterns, because, since electrons are indeed particles, they are also waves. The real shock comes when you only shoot one electron (or other particle) at a time through a sheild to create a pattern on film. Even though there was nothing for the particles to interfere with when shot one at a time, they *still* created a diffraction pattern.
This gives rise to the thought that particles that store their energy in 'quanta' and are small enough not to interact instantly with their environment, exist in multiple probability states. The electrons that created the diffraction pattern were interfering with the possibility that they existed elsewhere in the experiment.
In quantum computing, this is useful because electrons can be made to do different things at the same time, such as be in different places or aborb and release different amounts of energy. They can also simply stop existing at one place and start existing at another. They can also rock back and forth through time. Quantum computing, for the uninitiated, relies on harnessing these seemingly paradoxical phenomena. If the theories are all correct, this means that information storage will simply become infinite because there are an infinite number of states that any electron can occupy. Energy required to run a quantum process will be very little or zero, due to basic laws of thermodynamics and quantum physics. Speed of computations will be astronomical because quantum interactions take place on the pico-scale.
Quite a nifty thing...
Schroedinger's Cat says: It is not the world that must bend, but your mind. You must realize taht there is no mouse.
The next Slashdot story will be ready soon, but subscribers can beat the rush and slashdot the links early!
I suggest any reader interested in getting a good introduction to QC to take a look at a presentation given by Rob Pike at USENIX, available in MP3 audio here. It talks about the motivations on using information quantum mechanicaly (intrinsic parallelism, we are running out of atoms, etc); some historic aspects (Feynman's question: Can a computer simulate a QM system?); the approximations that you eliminate when you use QM computing devices (bits are not independent, but entangled); some algorithms (factoring, searching), etc. Not only nice, but funny too. Don't forget to get the slides also.
Just notice that there are two different aspects when we talk about QM systems, which most of the time are treated together: First, there is the QM way of representing information, which is to some point a reality now (on modern, high density Hard-Disk, for instance), the other is QM computers, which is something for way in to the future.
The simple answer is "possibly"
For example, it is possible that quantum computing can greatly increase 3D rendering. Basically, the main problem in ray tracing is finding the correct number of solutions that will lead a light ray to the point the eye is looking at. There are stochastic methods, like Metropolis, that greatly speed up the process of determining these solutions, but like most stocastic methods when compared to quantum methods, they are unreliable and slow (although when compared to deterministic methods, they are unreliable and fast). In a quantum 3D chip, you can theoretically easily find all of the solutions in a very short time, and thus determine the light levels for the point. This would in effect give you a perfect ray trace in a few cycles/point.
And even then, given enough qbits, you could be running those raytracing calculations on all of the points, oversampled by 256 to give a nice antialias.
But this is all in theory, because there are severe limitations on the logic that one can do with a quantum computers today. While the above could be modeled, I don't think we'll know for a while if it can be.
-no broken link