Using an A.I to train yourself to become a chess grandmaster
Image credit: Photo by Recal Media from Pexels
Contrived _voiceJan 30, 2022
Please leave the feedback on this idea
Is it original or innovative?
Is it feasible?
Is it targeting an unsolved problem?
Is it concisely described?
Bounty for the best solution
Provide a bounty for the best solution
Bounties attract serious brainpower to the challenge.
I used to play chess but quit when I noticed I wasn't the best at it. However, I also wasn't the best at art but I didn't quit doing it. That got me thinking about what was the difference between the two and I came up with the following reasons.
A lack of predictability in progress. Beating a person at chess doesn't mean you're better at it at them all the time. There is a factor of chance and circumstance at play. You may beat a player right now only for them to destroy you 2 hours later.
Frustration. Playing chess countless times and losing doesn't mean you're not getting better at it, everyone else could just be playing at a level you haven't reached yet. Even knowing that fact doesn't stop your brain from creating a negative emotional response to the game.
No precise feedback. After a game, an opponent might tell you your defense was weak or your attack was sloppy but no one really tells you how that was.
Differences in cognitive functions . This here is a game changer. You may emulate a really skilled player, analyse all their games and try to mirror their play style, However, In an actual game your perception of events and judgements on choice and consequence are made by your own mind. The difference in structure ends up throwing you further off your game.
While shifting to digital art I noticed something. Using my old works for reference produced significantly greater results in significantly shorter timeframes. Then It hit me. The only way to get better at something is to know EXACTLY what you INDIVIDUALY were doing wrong to begin with. Could The same be done with chess? I found a possible solution.
At first i thought playing chess against a computer would work , so I began research on how exactlt computers play chess. Chess is played on what is called a Thinking machine , it works by tracking the positon of each piece on the board along multiple possible dimensions ,then using that data it begins the task of increasing the value of it's pieces relative to it's opponents' while simultaniously taking control of the largest areas of the board. I found a website (thinking machine 6) that let's you watch it "think" while you play against it via transparent intelligence. However, not being a machine I could not find away to impove based on just raw data.
AlphaZero. A deep learning algorithm that you can train train to play chess. It is famous for teaching itself to play chess in 4 hours and recreating Gary Kasparovs open and dynamic style of play.
HOW IT CAN WORK
As an introduction it could play watch you play a pre-agreeed number of games against a thinking machine. Using it's observations it can now play against you as you allowing you to watch yourself mess up in first person. While you are busy beating yourself ,it records your new tactics and after a while uses them against you. It could be set to rotate tacticts after a week meaning you could always be playiing against yourself from the past week. This would let you see if you patched up the holes in your gameplay as well as adress new gaps that may have began to form. As you progress and keep beating yourself , your mistakes will become obvious to you since you will become familiar with how you think.
Once the A.I recognizes your play patterns as proffesional it will integrate other tecniques from professional players into its play and repeat the process. This could continue untill you were satisfied enough with your progress to join tournaments.The A.i could lead you on this path of defeating yourself for improvement until your play patterns were on the same level as the masters.
Although Alphazero is not open source, I did find Alpha-zero-general , an open source implimentation of Alphazero based on the AlphaGozero paper . It is supposed to work with any two player game. and is written in python.
I only have basic knowledge in programming however. Could such a system be feasibly made?
Human Parietal Cortex Structure Predicts Individual Differences in Perceptual Rivalry. Ryota Kanai, Bahador Bahrami and Geraint Rees in Current Biology, Vol. 20, No. 18, pages 1626–1630; August 19, 2010