Somebody in sci.physics once argued that a 'quantum computah' could
solve the traveling salesman's problem in next to no time.
'Somebody' in the unmoderated group like sci.physics is
not the Quantum Computing community. I'm almost sure
you have talked to some amateur which has just read an
article written by some popular science journalist without
a clue.
I replied that if you take into account that noise, or translated to moremundane speak,
the error rate of one quantum operation (25 % or so),
and then do for example 1000 such operations, the probability of you 'solving'
ANYTHING is less then the universe changing into a monkey as saying hello
Yes of course. That is very well known by the QC community, but
it may not be understood by an anonymous person posting to
sci.physics.
Agreed, there have not been that impressive results from Quantum
Computers, but the reason is not that those guys wouldn't understand
what noise is or what Shannon theorem says. The reason is that
the task is HARD.
You seem to be talking about quantum-gate based computers.
The technique perhaps more likely to give some practical results
soon is the Adiabatic Quantum computing, although it is
not a "full" quantum computer. It just finds global minima in
various optimization tasks.
I was a couple of weeks ago in a conference in Portland, where
Richard Harris from D-Wave explained their AQC hardware. Their
AQC finds the global minimum of a N-dimensional Ising system
with freely programmable couplings. Harris claimed that many
problems can be cast in the form of Ising system minimization,
including image recognition problems tackled by Google Inc.
Their computation time scales as
T[us] =5.84 N^2 + 65.5 N + 2E6 the function of the problem
dimensionality N. This scaling does not come from a theory,
but from the actual engineering of their functioning 128-qubit
processor, the D-Wave One. For instance, the constant
2-second -term comes from the time it takes the system to
cool back to the 20 millikelvin operating temperature, after
the couplings are programmed through RSFQ circuitry. The
programming cycle heats the system to 200 millikelvins.
Lockheed Martin bought one D-Wave One system
recently, I suppose not because it is practical, but because
they want to stay at the edge of the developments in the
field.
The classical supercomputers still perform the Ising
system minimization faster then the D-Wave One,
but the classical computational time increases ~exp(N),
when using the fastest known Simulated Annealing or
Iterated Taboo Search algorithms. Harris claimed that
the break-even complexity is roughly N~2000, and that
they are just about to roll out their 512-qubit
D-Wave Two processor.
The stuff above are from Harris' talk, it sounds plausible
to me, but I haven't studied all the details.All that stuff
is pretty new to me, D-Wave has been criticized in the
past about not telling the details of their work.
The rest of the drivel would take too long to comment,
here's just one thing:
Electronics has brought us many fun things, not that we really need those,
but usable in most cases, while all that crap physics has brought us nothing.
You don't count Bardeen's work as physics?
Regards,
Mikko
