AI and gaming share a rich, intertwined history. Beginning in the mid-twentieth century with the rise of computers, the challenges of programming machines to excel at intricate board games sparked considerable advancements in computer science. Chess, Checkers, Go, and Othello are a few instances where AI’s involvement transformed gameplay. This article will delve into this history, highlighting significant milestones and the emergence of AI game-champions.
The Genesis of AI in Games: Arthur Samuel’s Checkers
Arthur Samuel’s 1959 checkers program marked a significant early stride in incorporating AI into gaming. The program employed a system of fifty heuristics, and tested its prowess against differing versions of itself. The losing variant in each series would adapt the strategies of the winner. While this approach helped the program play robust checkers, complete mastery of the game remained elusive.
Chess and AI: The Long-standing Fascination
Chess, a game widely revered for its need for intelligence, has attracted AI development efforts for centuries. The first real chess program was developed in 1959 by Newell, Simon, and Shaw, following the Shannon-Turing Paradigm. Later, Richard Greenblatt’s program became the first capable of playing club-level chess.
Chess programs witnessed consistent improvement throughout the 1970s, achieving expert-level competence by the decade’s close. Ken Thompson’s Belle, developed in 1983, was the first program to reach the Master level, with Carnegie-Mellon University’s Hitech quickly following as the first Senior Master-rated program. This progression paved the way for Deep Thought, the first AI capable of routinely defeating Grandmasters.
Deep Blue: The Grandmaster Conqueror
Deep Thought transitioned into Deep Blue when IBM undertook the project in the 1990s. The software had its first face-off with World Chess Champion Garry Kasparov in a 1996 six-game match, which Kasparov won. However, a year later, against Deeper Blue – Deep Blue’s successor – Kasparov suffered a loss that sent ripples across the chess world. Even though these AI programs occasionally lost to top human players, their gameplay bore remarkable resemblance to that of skilled humans, indicating a crucial step in AI development.
Checkers Redemption: Jonathan Schaeffer’s Chinook
Jonathan Schaeffer, from the University of Alberta, targeted the game of checkers for AI domination in 1989, with his program, Chinook. Chinook faced Checkers World Champion Marion Tinsley in a 40-game match in 1992, losing only four games. Since then, Schaeffer’s team has been focusing on solving checkers from both the game’s end and beginning.
AI techniques have found application in various other games, including backgammon, poker, bridge, Othello, and Go.
Google’s AlphaZero: A Revolutionary AI
Google’s AlphaZero marked a milestone in AI’s role in games. An upgrade of Alpha Go, AlphaZero utilized self-play to master games, notably surpassing its predecessor in Go.
In a feat of AI learning, AlphaZero, having assimilated the rules of chess, trained itself to world-class level within a day. Astoundingly, it also developed a unique chess strategy that defied conventional wisdom and previous AI approaches.
While AlphaZero’s inability to explain its superior strategy raises questions, its self-taught proficiency undeniably showcases an extraordinary leap in AI game technology. So, does AlphaZero’s supreme ranking and self-learning qualify it as intelligent? The gaming world awaits an answer.
Related Articles
