As the development of technology, different kinds of AI begin to emerge and advance in the field of chess. In 2017, AlphaZero was invented and it gradually got famous for its unbelievable capacity of learning, which is called “deep learning”. It used only 9 hours to master chess by playing millions of matches with itself instead of learning from the games played by humans. The program generates 44 million game data after 4 hours of training, defeating top engines such as Stockfish.
Another kind of AI is created to help players to improve and make progress. Differ to AlphaZero, the AI “Allie” trained itself with the games human played (91 million transcripts from the popular chess platform “Lichess”). Thus, Allie plays similarly to humans and can adapt to various strengths, from beginner to expert. Moreover, this taught it how to make moves a human player would make, take time to contemplate critical positions, and even resign when the game is unwinnable. Unlike other AI which only wants to win the game, Allie can switch its strength when facing different people to make the games worth competing.
Additionally, researchers from Google DeepMind have created an artificial intelligence system capable of generating creative chess puzzles, which are strongly creative and imaginative. This makes the puzzles unique, interesting but hard to solve. Levitt, a British international master, who has written well-known books, mainly studying the thinking, intuition, and creativity of chess players described the project as a "pioneering step," saying: "This nature of collaboration is evolving, with AI now capable of generating interesting chess positions, beyond just 'mining' databases. The positions in this booklet represent a pioneering step in this human-AI partnership. While these initial AI-generated endgame compositions are not yet at a prize-winning level, they clearly demonstrate the potential to be."
What do AI bring to chess, better training ormalicious cheating?
To be simple, cheating with AI generally means using AI chess engines, such as Stockfish, to assist in decision-making during games, and use algorithms to generate optimal moves to replace their own chess skills, resulting in unfair competitive advantages. Nowadays, there are two main ways to determine whether a person is using AI. First, analyzing the consistency between real moves and AI engine moves is the most basic way, since the core trace of AI cheating is that the moves highly match with AI engine. According to the matches between grand masters (GM), the coincidence rate with AI is usually between 70% and 85%, if anyone can exceed this rate, you should suspect this person of cheating. Another method is to measure if the movements lack human traits. To be specific, even top-level chess players in humans may make calculation errors and omissions, but cheaters will avoid all low-level errors and frequently adopt moves which require deep calculation to discover.
However, AI can not only be used to cheat but also can help us improve. Artificial intelligence driven chess training tools provide players with targeted, human consistent methods to achieve tangible growth: Allie has been trained in 91 million human Lichess games, adjusting its strength to match beginners through experts, mimicking human thinking rhythms (considering key positions, abandoning unbeatable games), and revealing weaknesses in human style, allowing players to practice realistic, competitive scenarios. DeepMind's AI can create unique and imaginative puzzles, helping to broaden the thinking of chess players and cultivate their ability to creatively solve problems. By extracting human chess wisdom, it provides exercises that train tactical awareness, logical decision-making, and real game scenarios. These skills will directly translate into effective competition with human opponents.
Fairness in Chess
Fairness in chess can be understood as two players playing against each other without any help from others or other chess algorithm tools. When considering fairness in chess, we often think of output fairness and experience fairness. To be specific, output fairness outlines the reaction time differences between human and AI. While humans need to adjust their thinking time under different conditions during a match (affected by their feeling, the complexity of the situation, the energy loss caused by thinking and concentrating deeply), AI can output the best move in a short period of time consistently. As for experience fairness, it means that compared to humans learning their opponents’ habits in the games and studying how to play through thousands of chess records, AI can learn the opponent's moves and update their algorithm during the game and can learn by enumerating each possible move. Obviously, these make it extremely unfair to use AI in chess games.
How do platforms regulate these cheating behaviors?
Addressing these phenomena of cheating using AI, platforms like “chess.com” adopted the method of monitoring players who have extreme high win rates or have been reported by others for many times. If the system identifies the player is cheating by comparing the players’ moves with the AI model, their account will be banned, and they will not be allowed to sign un for another account on the platform again. Those who lost games against these cheaters will get their score of the games back to build a better environment of playing chess.
How do cheaters evade system monitoring?
To make the system harder to detect the cheaters, they will not follow every move of the AI. Instead, they may use AI to help them examine whether their moves will put them in an inferior position or not. Although this can decrease the overlap rate with AI moves, it will expose the typical characteristics of these cheaters: making almost no mistakes in their games. This feature is now gradually starting to be considered by the AI detectors, making the cheaters “impossible to survive”.
Future of monitoring AI usage in chess
Due to the trait of quick response of AI and the high accuracy of the moves in games, nowadays, we often measure the chess player's playing speed and accuracy under different levels of game complexity for evaluation. However, some cheaters will deliberately slow their playing rates down according to the game. This will lead to the problem that these detections are no longer useful on these kinds of cheaters. In the future, another method of monitoring AI usage in chess is to install local AI detection software on players’ devices. This can prevent them from using AI on the same device during the games. Nevertheless, if they try to use two devices to cheat, the detection will be much harder. By comparing their moves with AI and measuring their rate of bad moves are probably the most effective and efficient way to judge whether a player is cheating or not. To be simple, promoting the fair and reasonable use of AI in competitive games like chess remains a significant challenge for us.