I was crazy once…
… then again…
… and a third time…
… here we go again!
Concept
As an AI improves, using time efficiently becomes more and more of a challenge. People want more processing time, but we don’t want games to take forever to play out!
Proposal
This one is pretty simple. Run a contest with the bots’ time allotment doled out like a chess clock. Each AI is given a total allotment of processing time. The AI is free to divide this allotment between turns as it sees fit.
If the AI sees an opportunity to end the game early, it can chug on its computations a bit longer to get things right, and then deal the perfect move.
Or if it sees an obvious good choice immediately on a given turn, then it can send a response quickly, saving its CPU cycles for later.
If either side runs out of allotted computation time, then they lose.
- danBhentschel