All models are by their nature incomplete. They usually start out simple and become more sophisticated. It is prudent to defer policy decisions until one has some confidence in the models; nonetheless, right now, the modelling and the data are good enough to make a reasonably confident prediction about the magnitude of anthropogenic climate forcing in the next 50 years. We then need to look cool-headedly at the outcome, and decide if it's one we can live with, and, if it isn't, what we can do to ameliorate things. But arguing that because the models fifteen years ago were much worse, that they can't be trusted now, is foolish.
I disagree. The models support a conclusion, but have yet to be correctly correlated to historical data (to my knowledge). The issue is the bias in the model supporting a desired conclusion... That is the elephant hiding in the living room.