Tencent Wukong AI defeated professional players in an exhibition match during the semi-finals of the “Honour of Kings” World Champion Cup (also known as KCC). Tencent Wukong AI applies Tencent AI Lab's latest research in strategic and collaborative AI in the popular Multiplayer Online Battle Arena (MOBA) game, “Honour of Kings”. This victory advances Tencent Wukong AI to the next level and demonstrates its ability to solve more complex challenges at the level of professional players. 

“Honour of Kings” is considered to be an excellent experimental environment for testing AI's technical capabilities. The complexity of the game is 10 ^ 200000, and players have incomplete information. If AI can learn, analyze, understand, reason, and make decisions in real-time in such a complex environment, it can also be applied to changing and complex real-time environments. Many AI researchers in the industry believe that the next AI milestone is likely to be born in a complex strategy game.

The Tencent Wukong AI name implies “deep understanding and comprehension” and refers to a research project from Tencent AI Lab to explore new frontiers for strategic collaborative AI with deep intensive learning. The project, which began in 2017, has been successful in competitions with both top amateurs and professional human players. Tencent Wukong AI continues to evolve and build on its capabilities in long-term strategic insights, teamwork and collaboration. The insights from this research are designed for broader applications in the areas of eSports, robotics and other industries such as agriculture.

During the match with the professional players in the semi-final of World Champion Cup held in Kuala Lumpur, Malaysia, Tencent Wukong AI tested its capabilities with professional players. Tencent Wukong AI managed to pass the test with strategic moves and upgraded to the professional level within 30 minutes.

Following the test, Tencent AI Lab technicians shared the technical details of Tencent Wukong AI. Tencent AI Lab has established an algorithm with an enhanced learning model based on “observation-action-reward”. Tencent Wukong AI plays against itself and can train at the equivalent of more than 400 years of games per day, allowing it to learn basic techniques such as how to avoid skill damage. It can also explore new strategies which differ from what one usually observed in human players. 

“Honour of Kings” is a complex multiplayer strategy game, where players face a large number of potential actions related to strategic planning, hero selection, skill application, path exploration and teamwork. Tencent Wukong AI uses one model to create and control multiple agents, achieving teamwork and coordination comparable to professional players. Recently, Tencent Wukong AI’s one-on-one model achieved an average success rate of 99.8% playing mobile games with the general public at an event held in China. If AI can learn, analyze, understand, reason, and make decisions in real-time in such a complex environment, it has the potential to be applied to real-world problems with similar challenges.