Most of the feedback loops are positive. Clouds, a major factor, are uncertain. The ultimate effect of clouds is dependent not only on whether cloud cover increases or decreases, but on what kind of cloud cover increases or decreases (i.e., low cumulus or high cirrus, etc.)
Many clouds such as the cirrus clouds blown off from thunderstorms to the west of me, are not in the model. They are cooling my locale. Cooling in any locale without corresponding warming in another will result in model bias. Models are not so much "uncertain" as they are biased towards warming. The only way to believe that increased water vapor does not cause more cooling clouds is to leave them out of the climate model.
Nonsense. Systems where positive feedback dominate blow up.