Two articles on global warming computer models. The first is a lucid explaination of how the climate models work, and why it is impossible to verify the models.
The second one is on how the modelers themselves have it impossible for the models to be validated (a usually necessary step in determining how well models actually work).
Article #1: The Trouble with Global Climate Models
Article #2: Validation Of A Climate Model Is Mandatory: The Invaluable work of Dr. Vincent Gray
Fudging and Nudging
The climate models cannot be validated because they should be set up so that they accurately predict the past ("hindcasting"). However, the models cannot do so until the modelers have added a number of fudge factors that nudge the curves into the right shape BEFORE they run it. So, of course, it looks like the models are doing what they are supposed to.
To put this into computer programmer terms.
Imagine a computer programmer programming a computer that runs a robot that makes light bulbs or wraps candies (I know people who do that).
Evaluation of the model is like the programmer giving the program to a systems analyst to check the program. The analyst may run parts of it in a simulation, but does not actually use it run the robots.
Validation of the models is like the programmer running the program and seeing if the robots do what they are supposed to do. If the robots correctly wrap candies, the program is valid.
The programmers of the climate models have made it impossible to actually validate their models by throwing in enough "fudge factors" to make the model work on the past ("hindcasting") that the models cannot predict future ("forecasting"). The fudge factors push the models to predict things that are not happening.