Change of centipawns for Computer Analysis

Recently we have seen the new computer generation playing chess, the famous Artificial Intelligence.
We have Neural Networks working quite great and nice, sometimes more strong than the best current computer engines.

We all know the "points" of each piece. The queen worh 9, the rock worth 5, knight and bishop worth 3 or 3.5; and pawns worth 1. This calculation is fair enough and we have all used somehow for our games.

With the new generation of Neural Networks, lots of this "points" we have all accepted during many years are being questioned. Aparently, AlphaZero rates more Bishops than Knights, and rates more the Queen than 2 rocks (at least, from what I have seen from their games)

My question is:
How Stockfish and other computer engines will be affected if we would change this punctuation of the pieces? Would they be stronger?

And what about real chess; do you think strong players will change somehow their way of playing if we would change the punctuation of the pieces? Of course its well known in theory what pieces are better and in which conditions, but I wonder if this new contribution could change somehow chess.

Thank you

You cannot do such conclusions from limitted set. Particulars of position dominate. Bishop is by and large better than Knigth but not much. But Knight on right place is worth more than rook. And a0 games are all about the particulars. On most beautiful ones it destroys the mobility of the opponent to a degree that it does not matter whic pieces it has . Piece not doing anything is not worth much