That’s an interesting idea.

But the magic is in the “find patterns” stuff as memorization is just data storage. If you think of the machine learning algorithms as assigning items a point in a space, then it does uncover neighbors, sometimes ones we might not expect, and that’s interesting for sure.

But I’m not sure it’s analogous to what people do when they uncover patterns.

Definitely interesting to ponder though.

No, because ML is compression via interpolation and does not imply decompression.

That really depends on the particular algorithm.