For inference, we have hand crank that rotates a lot of gears, with a final gear making one token (word) appear in a slot. For learning, we even know how to fee

For inference, we have hand crank that rotates a lot of gears, with a final gear making one token (word) appear in a slot. For learning, we even know how to feed a bunch of text into a complicated thing that tells us what gears to connect to each other and how. We have no idea why the gear ratios and placements are what they are.