Tuesday, January 22, 2008

A Model of Error

Over at UD, math (sic) professor Granville Sewell resurrects one of the silliest posts of 2006, Gil Dodgen's misapprehension of what makes a model. No, the errors in our OS are not part of the model, any more than they are part of your PDE solver or Dr Dembski's MESA or Dr Marks' EIL work.

It is however true that if you are running intense numerical calculations, you need to take into account the quality of your (pseudo)random number generator, your floating point hardware's aptitude for error due to heat, cosmic rays, bad programming by Intel, etc. Taking these things into account is very different from intending them to be part of the model system. No one is going to be impressed by results that need to be run on Windows ME with a lump of uranium lying on the desk next to the CPU.

But it is from this support the Dr Sewell makes a sweeping statement, "Unintelligent forces simply can’t do intelligent things..." Sorry, Dr Sewell. In the words of the prophet, pwned! May I be the first to introduce you to the Humies.

If you're not following the link, I'll just summarize that the Humies are cash based awards for achievement equal or better than human at some intellectual task, by an EA based system. That we can give away awards like this is a testament to the power that GA, GP, etc have already achieved.

No comments: