Applying all chess principles in chess without mistake, how far does that get you?

@15

"How would it "know" what a loss of material even means?"
++Just tell it N = B = 3 P, R = 5 P, Q = 9 P, K = 1000 P
You can refine it, or it can find itself by playing against itself.
AlphaZero arrived at P = 1, N = 3.05, B = 3.33, R = 5.63, Q = 9.5

"incorporate some sort of calculative ability" ++ Yes. The engine can play 2000+ on just calculation and simple evaluation. It cannot even reach 2000 without calculation.

"It needs to have a way to assess the position"
++ Just counting material is enough for an engine to play 2000. A weak pawn is then only considered weak if it is lost within x moves where x < horizon of calculation.

"once we give the program the ability to calculate, then we're no longer relying on chess principles"
++ A program cannot rely on principles alone, but it can rely on calculation alone. In the initial position or other quiet positions no calculation can really result in what move to play. On the other hand in quiet positions there are several good moves and it usually does not harm to play a subpar move.