I'm trying to develop a better understanding of the concept of "out-of-distribution" (generalization) in the context of Bengio's "Moving from System 1 DL to System 2 DL" and the concept of "(meta)-transfer learning" in general.
These concepts seem to be very strongly related, maybe even almost referring to the same thing. So, what are similarities and differences between these two concepts? Do these expressions refer to the same thing? If the concepts are to be differentiated from each other, what differentiates the one concept from the other and how do the concepts relate?