I programmed computers for over 20 years. The accuracy of any computer model is based on how well the underlying system being modeled is understood. Climate is proving more complicated than many people first assumed. Some earlier models didn't even factor in cloud cover. Garbage in, garbage out applies to climate models just as it does to all other forms of computer software.

At present, I don't know of a single climate model that has been validated by comparing it's predictions to real-world observations. This involves starting at some known conditions at some time, say 100 or 1000 years in the past, and let the model propagate to the present day and compare the predictions to what actually happened. Until a model has been validated against the real world, it's little better than a hypothesis. In the past, when I've pointed out that the latest touted model needs to be validated, I was told there wasn't time. As my father always said, "If you don't have time to do it right, when are you going to have the time to do it over."