Here is my thinking. According to various algorithmic/probabilistic accounts of "it" from "bit" and such, abduction makes no sense from a computational perspective. The simplest, most efficient explanation for the universe is that appearance IS reality. ML is pure empiricistic induction. If you're asking why deep learning hasn't figured this out yet it's because, probabilistically, it's more likely that a DNN popped out of thin air compared to the probability a thing called a "human" popped out of thin air then built a DNN. If all we have are random collisions of particles, the 1st is much more likely to occur in the history of the universe. This suggests to me humans are somehow hardwired to assume a distinction between appearance and reality--a transcendent realm Kant told us was off-limits.