Exactly :) In most cases, unsupervised learning is nothing more than having the AI try to approximate the solution of your highly non-linear loss function. So

Exactly :) In most cases, unsupervised learning is nothing more than having the AI try to approximate the solution of your highly non-linear loss function. So if there's any way of solving that loss function directly, it will perform like a well-trained AI.