Have you ever stumbled across a random YouTube video where a robot is playing chess with a human? Well, even if you manually searched for the video, we won’t judge you because this is how fascinating these games are to watch.
A Quick Flashback Into The History Of Chess And AI
Back in 1890, a Spanish scientist invented a system that could checkmate the opponent in a king and rook endgame. However, the problem was that the system did not always make the best moves. Instead, it would play about 50 moves before finally delivering a checkmate. To get an idea, it should take less than 20 moves to give checkmate in such an endgame.
Although there was no practical purpose of such a device, it was a significant push towards programming machines in such a way that they make legal moves. After this, however, there was no vital progress in the field for a very long time. In 1947, Alan Turing trained a machine to analyze one side moves. This was the moment when the first baby steps were taken towards leveraging emerging technology.
Just a year later, American scientist, Claude Shannon, secured a significant accomplishment through his groundbreaking paper entitled Programming a Computer for Playing Chess. The report described how machines could be programmed to play chess based on an evaluation function of a given position on the board. This paper was a crucial step towards automating chess. The evaluation function was vital in order to find and assess thousands and thousands of different positions. Ideally, about 27,000 different positions are examined by the engine to account for both Black and White moves. However, it has been estimated that humans can calculate only about 50 positions while making a move. And mind you, we are talking about the Grand Masters of chess.
In the 1960s, computers began competing against humans. In fact, in one of the tournaments, a program written by an MIT graduate drew one game with a human. Over time, the performance of computers increased considerably. Several factors were considered to refine the evaluation process. In the ’70s and ’80s, technological advancements fuelled the progress in the automation of chess. By the late 1980s, computers were able to defeat several top-notch players in important tournaments. These computers were able to assess 175,000 to 700,000 positions per second.
Over the next ten years, computers such as Deep Blue were able to assess as many as 50 billion positions in about three minutes. In 1996, Deep Blue defeated the World No. 1 of that time, Garry Kimovich Kasparov, in a series of 6 matches.
However, this was met with criticism by some players who thought the game had deteriorated in terms of creativity due to the soulless calculations performed by computers.
Did AI Wreck The Game?
Russian chess grandmaster, Vladimir Kramnik, talking about the automation of chess, stated:
After Kasparov’s defeat in 1996, Bobby Fischer, a renowned player, called for a meeting to revamp the entire game in order to push creativity. He introduced the Fischer Random Chess, also popularly known as Chess960. In this format, the initial position of the pieces is randomized. All the other rules of the game are retained as it is. This format gained a lot of popularity over time.
Learning A Game From Scratch And Adding Creativity
In 2017, a UK based artificial intelligence company DeepMind introduced AlphaZero that mastered the game from scratch. The system could learn the game from scratch to discover new variants much quicker than humans ever could. A team member of DeepMind, Tomašev, stated:
AlphaZero tested nine different patterns in the game. This included eliminating the castling of the king. In five more variations, the system modified the movement of pawns, including a variant called Torpedo Chess in which pawns can take up to two steps at any time during the entire game, instead of the traditional game format where pawns are allowed to take two steps only in the first move.
Kramnik, the brain of the project, felt that the beauty of the game was restored again with the transition to new variations. For instance, the elimination of castling created rich new patterns in order to protect the king. Another variant where players could capture their own pieces to move ahead enticed the grandmaster. He stated, “All in all it just makes the game more beautiful.”
Even as players continue to rely on computers and deep analysis to improve their game, this fresh application of AI is expected to bring in more traction to the game.
Machines have improved to such an extent that beating them is nearly impossible. In such a scenario, researchers are using AI to push creativity in the game instead of using the technology to defeat humans.