The "knowing what it doesn't know" problem is trickier even than it might seem, too: right now we don't even know how what it does know is encoded in its neuron

The "knowing what it doesn't know" problem is trickier even than it might seem, too: right now we don't even know how what it does know is encoded in its neurons, other than some vague notion that the knowledge is somehow "distributed" across a ton of distinct neurons (there is no specific neuron/set of neurons that are dedicated to memorizing X or Y fact), so figuring out what it doesn't is not trivial at all