Picture these staggering fast processing speeds applied to genetic algorithms/programming. For those who don't know, in a very rough sense (don't nitpick) this involves:
1. You want a program that take input X and produces output Y
2. Generate a whole bunch of random programs. Most will do nothing even resembling producing the output Y, but some will suck more than others.
3. Feed the input X, and see what output each program produces. Take the least terrible program from the batch, make a bunch of copies
4. Then for each copy, change it randomly a little, eg flip a few bits.
5. If there isn't any program in the bunch that's good enough for you, go back to 3.
6. Otherwise, you have your program and didn't have to write it yourself. In fact, you don't even need to have the slightest idea how to solve the problem, just how to state it.
The only problem with this is that it may take a really, really long time to evolve an acceptable program. Often too long to be worth bothering. But with speeds as ridiculous as they propose for quantum computers, what program couldn't you evolve in say, a day? Or, for that matter, why not just generate every possible piece of machine code of a given length, run them all through an emulator at sickening speeds, and see if any of them solved the problem? I think that if you have truly sick processing power like this, then almost any problem is solvable with relative ease. Maybe I overstate the case, but you see what I'm getting at.
Slashdot Mirror
Picture these staggering fast processing speeds applied to genetic algorithms/programming. For those who don't know, in a very rough sense (don't nitpick) this involves:
1. You want a program that take input X and produces output Y
2. Generate a whole bunch of random programs. Most will do nothing even resembling producing the output Y, but some will suck more than others.
3. Feed the input X, and see what output each program produces. Take the least terrible program from the batch, make a bunch of copies
4. Then for each copy, change it randomly a little, eg flip a few bits.
5. If there isn't any program in the bunch that's good enough for you, go back to 3.
6. Otherwise, you have your program and didn't have to write it yourself. In fact, you don't even need to have the slightest idea how to solve the problem, just how to state it.
The only problem with this is that it may take a really, really long time to evolve an acceptable program. Often too long to be worth bothering. But with speeds as ridiculous as they propose for quantum computers, what program couldn't you evolve in say, a day? Or, for that matter, why not just generate every possible piece of machine code of a given length, run them all through an emulator at sickening speeds, and see if any of them solved the problem? I think that if you have truly sick processing power like this, then almost any problem is solvable with relative ease. Maybe I overstate the case, but you see what I'm getting at.