When I was an undergrad a long time ago, I implemented a backgammon AI as a course project using minimax. Of course, neural networks etc. were available, but all methods at the time relied on the fact that there were some programmer-defined metrics which could be measured, and assigned weight to each metric to evaluate alternatives.

I have been a programming languages researcher, mostly away from AI, so I would like to know what the state-of-the-art is today. Is it possible to use commonly available tooling to only define what the rules of the game are and what each move does, and let the AI tools determine the metrics and their weights automatically?

What is the quickest and ideal way to implement a backgammon AI today? I am looking to do it in order to get a sense of what is possible today, and what the tools allow.

Thanks!