I get the most satisfaction out of building small solutions that work really well. Problems that are beyond the scope of an individual to solve are not really as motivating. At that point you are just one of a group who slings code at the problem and hopes some of it sticks.

Exactly. Or you can go physical and do some HW+SW project, there is no AI (yet) for the physical world, at least at a price point for hobbyists. Although you can already go a long way 3D-printing