How big to avoid AIs?

Simple way to think about it:

imagine that 9k who always played on 9x9 only, first time plays 19x19
his power possibly will be 15k
And what if there is 9D 19x19 player who can tell each time in which corner to play?
9k would play definitely better than 15k with that cheat.

We can train AI on 10x10 board with special border properties
then we can train AI that only able to choose 1 of 4 possible options,
small AI plays locally automatically, big AI don’t know about existence of small AI, small AI is part of rules of game that big AI plays.

Doing the same computation, but this time with upper limits for computing systems. There is a theoretical limit for the number of computations a universe sized computer can do: roughly 2.8 × 10^229 operations.

Let’s assume the number of computations per move is at least equal to the square of the board size S (although a quantum computer may be more efficient than that?). The number of moves in a game is roughly S squared.

So a theoretical limit of S for a computer the size of the universe would be the fourth root of 2.8 × 10^229 operations = 3.0 x 10^57.
This AI would not be able to finish a game on a 10^58x10^58 board in its lifetime of 10^132 years.

Edit: fixed calculation error

How disappointing!

1 Like