I said they have to be validated. Computer models are only as good as our understanding of the fundamental forces at work in the system being modeled. Validation is essential for checking our understanding. An example was what happened in Computational Fluid Dynamics back in the 1980s and 1990s. The Pegasus launch vehicle was the first to be designed completely using CFD. NASA sponsored a payload on the first launch to validate the accuracy of the model predictions against the real world. They did other experiments on the Shuttle as well.

A model that hasn't been validated is little better than a hypothesis - not completely useless but not yet trustworthy. How well do the climate models cover things like solar variablility? Global warming has been observed on Mars, and since I doubt that's due to human activity, it seems the common factor is solar output.

As good as the hurricane models are (and they're the result of over a decade of work based on actual storm data), none of them predicted what happened with this one. Within 12 hours of landfall, it weakened from a Cat 5 to a strong 4 and shifted to the east. The real world can be a complicated place.