i feel like a lot of applications of ai basically entail giving up on the underlying problem and hoping we can outsource the tedious and unpleasant part of wading through it to ai
kind of like if we decided that because cities don't do a good job plowing roads, we're just going to have everyone buy huge trucks.
ofc, sometimes the right (or "right") solution is intractable or hasn't been developed yet so we're better off implementing stopgaps or other solutions that don't necessarily touch the root cause. i think where the latest round of ai fails is that it's often expensive and unreliable, and we often do have the right solution. also ai is just kind of being thrown at these problems without much understanding.