The Pros and Cons of Generalization in Intelligence

Instead of simplicity I'm using this thing called weakness and so I've got this this proof that weakness is optimal but if I'm trying to predict language. So I take the set of choices which would explain the behavior I'm seeing based on my own preferences and then search for the weakest hypothesis within that. Something that you know contradicts the fewest number of possibilities could you contrast maybe the motivation for developing weakness? If we want a practical usable version we'd like some computable version of it what limits can we impose that get us there, he asks.