With a technique known as genetic algorithms it's possible to solve a problem by:
Generating a random set of solutions (the "population")
Creating the next generation by randomly combining the best solutions in the current population
Repeating the previous step until you are happy with the current best solution
Genetic algorithms have been used to solve a number of problems that would be hard to solve otherwise, problems where it is easy/fast to determine if a solution is any good or not.
If we combine this with test-first development, where you start writing testcases, then code until they pass, we would end up with a software development methodology where all the developers do is write testcases, then GA will provide a solution where all the tests pass.
Think about it, we have loads of processing power available so who cares if you have to run your tests every night on a huge system - when you show up in the morning you are likely to have a working solution that passes your tests (until you write more tests, that is).
Mind you, GA tends to write absolutely horrible code, and most of it by far doesn't even seem to be used at all - which is also the case with DNA, BTW. And for a complex system you need a huge number of testcases. Maybe the issue in the future will be how to automatically write tests?
Slashdot Mirror
Genetic algorithms have been used to solve a number of problems that would be hard to solve otherwise, problems where it is easy/fast to determine if a solution is any good or not.
If we combine this with test-first development, where you start writing testcases, then code until they pass, we would end up with a software development methodology where all the developers do is write testcases, then GA will provide a solution where all the tests pass.
Think about it, we have loads of processing power available so who cares if you have to run your tests every night on a huge system - when you show up in the morning you are likely to have a working solution that passes your tests (until you write more tests, that is).
Mind you, GA tends to write absolutely horrible code, and most of it by far doesn't even seem to be used at all - which is also the case with DNA, BTW. And for a complex system you need a huge number of testcases. Maybe the issue in the future will be how to automatically write tests?
</incoherent-rambling>If a Zen Master is offended, and there's nobody around, will his MP3 player still make a sound?