The AI chess coach

A robot developed by Taiwan engineers moves chess pieces on a board against an opponent, ,at the 2017 Consumer Electronic Show (CES) in Las Vegas, Nevada on January 8, 2017. The robot developed by Taiwan’s Industrial Technology Research Institute, which spent the week playing games against opponents at the Consumer Electronics Show, was displaying what developers call an “intelligent vision system” which can see its environment and act with greater precision than its peers. / AFP PHOTO / Rob Lever####################ROB LEVER

In a world increasingly taken over by robots and AI, it is not inconceivable that the job of chess coaching can be taken over by machines too.

We have all seen what the latest machine learning technology in Alpha Zero has achieved. It beat grandmasters not by sheer brute force of calculations and computations like the technology used by Stockfish, Komodo, Houdini, Fire etc but by using machine learning.

Deep Learning

This is a game changer as the computer now learns by trial and error as well as real world data fed into it’s hungry data banks. This is called Deep Learning. Achievements in this field is numerous including autonomous driving cars, computer vision, studying customer behaviours etc. What is amazing in the case of chess is that AlphaZero is self learning.  In just a few hours of self-training, starting from just the basic rules of the games, it became strong enough to best the strongest engines in the world.

The same technology can be used in chess coaching. Beating humans and other engines is impressive but it would be a bigger achievement if the machine can teach!

AI Coach

It learns by observing the games of it’s student. It’s also fed with other real world games and positions which it starts to classify based on themes and motifs. At first it will fail miserably but as with Deep Learning technology, it quickly picks up.

It gets better at identifying weaknesses of the human student and prepares test positions and test games where it drills down and highlights the solution. It gets better at classifying games and position based on it’s own themes and what it thinks matches the needs of the students.

What we get is not just an AI chess coach but a coach that learns how to teach better as time goes by.

The different levels of a chess trainer as classified and certified by FIDE:

FIDE Senior Trainer (FST)
FIDE Trainer (FT)
 FIDE Instructor (FI)
 National Instructor (NI)
 Developmental Instructor (DI)

Imagine you have a DI (Developmental Instructor) level coach but slowly gets better and teaches even up to the level of FST!

Crowd sourcing multiple students data from different geographical areas, age, race, gender etc into a central database makes the teaching/learning process even more refined.

Sounds too futuristic and impossible? That’s what AlphaZero sounded like before it came into existence.

Leave a Reply

Your email address will not be published. Required fields are marked *