> A model that knows more in general, will often be better at specific tasks

Is this your hypothesis or broad conclusion among AI experts?