In the fast-changing world of gaming, artificial intelligence (AI) has become a powerful force, challenging and beating human champions in many games. From Chinook’s historic win in checkers in 1994 to OpenAI Five’s recent triumph over world champions in Dota2, AI’s dominance in human-computer gaming is clear.

Motivated by the goal of achieving Nash equilibrium and advancing general AI technology, researchers have explored different game genres, including board games, card games, first-person shooter games, and real-time strategy games. They have made significant progress in conquering complex games like Honor of Kings and Mahjong.

One of the most memorable moments in AI gaming history was Deep Blue’s victory over chess grandmaster Garry Kasparov in 1997. This marked a turning point in strategic decision-making and showcased AI’s immense potential. The paper aims to introduce newcomers to the field, discussing techniques, challenges, and notable AI systems that have contributed to human-computer gaming.

AlphaStar, an innovative AI, has reached grandmaster level in all three races in the complex real-time strategy game StarCraft. Its counterpart, Commander, a lighter version of AlphaStar, impressed audiences by defeating two grandmaster players in a live event. These achievements demonstrate AI’s exceptional ability to excel in competitive gaming environments.

The true test of AI performance lies in surpassing professional human players. Several AI systems, including AlphaGo, AlphaGo Zero, AlphaZero, Libratus, DeepStack, DouZero, Suphx, FTW, AlphaStar, OpenAI Five, JueWu1, and Commander, have undergone rigorous testing against top human players in various game genres.

The Turing test, a human-computer game designed to determine if a machine possesses human-like intelligence, is essential for evaluating AI systems. The paper explores the challenges faced by current techniques in human-computer gaming and discusses future trends. Tree search (TS) and distributed deep reinforcement learning (DDRL) are identified as key approaches in game AI development.

Games with incomplete information pose unique challenges for AI systems. Counterfactual regret minimization (CFR) has emerged as a leading algorithm to overcome these obstacles, as seen in DeepStack and Libratus’ victories against professional poker players. The paper highlights the importance of population-based training and population-play for creating scalable agents with different policy networks.

Throughout the article, the significance of computational resources in training high-level AIs is emphasized. Limited resources often hinder AI development, posing a challenge that needs to be overcome.

AI’s rise in human-computer gaming goes beyond specific games or maps. AIs like JueWu have successfully played full real-time strategy games, while Suphx has outperformed most top human players in Mahjong. These achievements showcase the versatility of AI systems and their potential to revolutionize the gaming experience.

In conclusion, the article stresses the need for further experimentation and exploration in game AI, particularly in introducing disruptions within the game environment. Overcoming these challenges will lead to AI systems reaching even greater heights in human-computer gaming.

The journey of AI in human-computer gaming has been truly remarkable. From defeating world champions to reaching professional human player levels in various games, AI has undeniably demonstrated its capabilities. This comprehensive overview of techniques, challenges, and opportunities in the field offers readers an exciting glimpse into the future of AI-powered gaming. Get ready for an exhilarating revolution in the world of human-computer gaming as AI continues to push the boundaries of what is possible.