I see two distinct cases here:
Imagine an infant saying: Well, it might be cool to learn to walk now. First, the child could never know where the ability to walk will take it. Second, the choice to learn things that transcend one’s horizons/knowledge cannot be based on what’s in it form me. An example of things that fall into this category for me was discovering erts/Elixir and AI employing genetic algorithms. The amount of need for exploratory learning and the inner drive behind it is highly individual, of course.
When I know the desired outcome, for me, it boils down to: what do I need to learn in order to achieve goal X? Once the knowledge seems unfit for fulfilling such goal: rinse and repeat. I believe it was in The Gay Science where Nietzsche talked about a savant who dedicated his life to the study of a part of a worm’s brain. When such knowledge is not practical nor expands one’s horizons and basically just uses up more memory, then, I would channel the energy elsewhere. It might be just another shape the procrastination took on.
Having such question, I am surprised you loathe neural networks-related knowledge (or is it just deep learning/Tensor Flow crap?). I would look into AI parts where fitness/reward function is mentioned. There you might get answers on a theoretical level.