An even better mental picture to push back on might be an “annotated” version of the singular learning theory toy picture, where by “annotated” I mean that you’ve put mental labels on the low-dimensional minima as “algorithm 1”, and the high-dimensional minima as “algorithm 2″. The minima don’t have to correspond to a single algorithm as we conceive it.
An even better mental picture to push back on might be an “annotated” version of the singular learning theory toy picture, where by “annotated” I mean that you’ve put mental labels on the low-dimensional minima as “algorithm 1”, and the high-dimensional minima as “algorithm 2″. The minima don’t have to correspond to a single algorithm as we conceive it.