Masayuki Mochizuki

Let me explain a little bit how these AI algorithms work.

You have, let's call it a play agent.

That's the agent making the moves.

And then you have a learner.

We call it the value function in machine learning.

So you're training your value function as a function of the results that the play agent decisions are making.

Basically, it's a self-learning algorithm.

The best move is simply the move that has the higher value in the value equation.

Because we have the computer engines, the neural nets that more or less solve the game, the level at the top is so high nowadays that it's much more a game of not making mistakes.

Does that take some of the fun out of it?

I don't think so.

If you are a really high level player, what is your motivation?

For me, it's this mastery achievement thing.

I'm not playing against you.

I'm playing against myself.

Here's how Bob Wachtell described it in his book, The Backgammon Chronicles, A Pro's Adventure on Tour, Volume One.

Backgammon has evolved from a model like that of soccer, basketball, or tennis, he wrote, where it is the fantastic inspired shot that will be the object of awe for years to come.

to a sport like gymnastics or ice skating or diving, where the entire goal is to perform a flawless routine, a perfect 10, i.e.

a zero error rating.

I went back to Wachtell to talk about this evolution.