Comment by Davidzheng

a large number of breakthroughs in AI are based on turning unsupervised learning into supervised learning (alphazero style MCTS as policy improvers are also like this). So the confusion is kind of intrinsic.