It is precisely because of the complexity of the climate that one cannot just plot CO_2 concentraions against temperature. One has to make detailed models and validate those. Now, I've read that climate models do an excellent job in describing past climate. Of course, one can object that one can always adjust parameters in such models to fit observations. But these models are based on physics and are thus well motivated.

If you leave out the physics and just look at correlations you can't conclude anything...