On: Can Neural Networks Crack Sudoku?

[1001101]

If the wet ones can, I don't see why the ones of an arbitrary size made out of bits can't. Not a very good argument, but there's this: recurrent neural networks are Turing complete.

[sp332 → 1001101]

Humans don't learn how to play Sudoku by staring at a bunch of solved puzzles. We are given the rules up front.