Wednesday, December 14, 2011
What Do I Think of Computer Models?
Computer models are buggy self-fulfilling prophecies with circular dependencies. I write software for a living, so I should know. Computer models can only model what we already know, so they can't actually teach us anything. They certainly are not useful as experiments to prove a hypothesis, because the computer model simply becomes the hypothesis -- implemented as a computer program. It's a tautology; nothing more. If you're actually trying to prove or disprove a hypothesis, you have to abandon virtual reality, and hie thee into actual reality, run real experiments, and start taking measurements. Only then will you learn anything new. On the other hand, if all you're trying to do is baffle politicians or the media with bull to promote an agenda, a computer model might do nicely.