"Enlightenment AI+ Game College Competition" Launched, Why Should AI Learn to Play king of Glory?
Enlightenment AI + Games Competition in Colleges and Universities
On August 18, Tencent announced the launch of the first "Enlightenment AI+ Game College Competition" preliminary contest, which is also the first time Tencent's Enlightenment platform is open to universities. The competition was jointly launched by Tencent AI Lab and King of Glory, Tencent University cooperation, Tencent Game Academy, etc. More than 100 students from 18 top universities in China, including Tsinghua University, Peking University and Shanghai Jiao Tong University, were invited to compete on the same stage.The participating teams need to train the optimal model under the given resources within a specified time, and finally deploy the AI server using the optimal model, and submit the system for 1V1 (preliminary) and 5V5 (final) matches. The prize pool is as high as 500,000 yuan and the final result will be released in December.The basic gameplay of Honor of Kings is to gain advantage by pushing towers, killing enemies, and earning money by destroying the enemy's crystal base.
AI Game Technology
Seemingly simple rules are a huge challenge for AI to make complex and fast decisions with incomplete information and high complexity. In large and incomplete information on the map, 10 participants in strategic planning, choice of heroes, application skills, path to explore, and team collaboration, a large number of uninterrupted, instant choice, which has led to a very complex situation, is expected to have up to 10 20000 kinds of operation possibility, and the whole universe, the total number of atoms is 10 to the 80 joules.To achieve the goal of victory, it is necessary to examine a number of AI capabilities: agent solutions, features, action space, model structure design, reinforcement learning algorithm design, reward function design, training mode exploration, etc.
Future Possibilities for AI
Wen Ying, a professor at Shanghai Jiao Tong University, has also proposed a new technology that allows AI to cooperate in more diverse ways, making it less likely to reveal their routines and making the game more challenging. The technology could also be used to train self-driving cars. Tencent AI Lab itself is also experienced in the field of intelligence research. Since 2016, the weiqi AI "Fine Art", which was developed, now serves as the special AI for the Training of China's National Go team. In 2017, based on the test environment of King of Glory game, MOBA strategy cooperative AI research and development was launched. In 2019, it reached the level of professional esports. Tencent also won VizDoom, the top AI competition in the shooting category, and was the first to develop an intelligence that beat the built-in AI in StarCraft ii. Can an AI that learns to play King of Glory learn to deal with more real-life problems? What are the possibilities for AI in the future?
