The difficulty in simulating any non-linear system like clouds, climate, and weather is that it will either tell you what you want to hear, or exactly what your empirical data points to. The system can literally do ANYTHING. The only option is to create an array of probable states and look for the one that is more likely than the others.

At a large scale we have some metastable states that are kinda synced to some large linear input. But for all our predictions, the math says we could be experiencing 200C by September, and -80C by December.

The only reason to believe that these states are unlikely is they haven't happened in the past. At the same time, there hasn't been a carbon dump into the atmosphere at this scale so consistently for the entirety of our recorded history. So we are kind of entering unknown territory.

I predict the predictions will remain very dire until December 2020. Then they will back off, probably with little note of explanation.

The predictions will again become shrill in February of 2024. Then they will back off in December.

The predictions will again become louder in February of 2028...

TLDR; Clouds reflect light depending on the amount and how high they are, this contribute to greenhouse effects. Apparently, this wasn't part of the models before.