Its a no secret that Tencent is involved in AI and related stuff. A month ago it was revealed that Tencent is also creating its virtual human. Now the company has revealed that they are using the massively popular game Honor of Kings for AI Reinforcement Learning that will solve real-life problems.
Tencent held a competition to find the best brains
Student developers worked hard to create AI algorithms based on reinforcement learning (RL) that may be used to play the Honor of Kings autonomously during the second AI Arena Multi-agent Reinforcement Learning Competition in China, which finished in April.
The winning team, made up of five Tsinghua University students, claimed that when used in practise, the theoretical capabilities of the RL model were more difficult than anticipated.
Smart solutions for complex problems
Computer game agents, like human players, develop greater intelligence when they encounter novel behaviours and consider the proper course of action. The underlying technology is called reinforcement learning (RL), a type of machine learning paradigm in which programmers reward actions they want the AI to do and the computer trains or learns by carrying out the required operations to get the intended behaviour or result.
In the past, deep RL algorithms were tested using board games like chess and go. AlphaGo, a computer software created by Google subsidiary DeepMind Technologies, is one of the most well-known instances. In 2016, AlphaGo defeated 18-time world champion Lee Sedol in a game of go, winning 4-1.
After six years, board games have been replaced by more sophisticated, imperfect information games and strategic video games as the primary focus of game AI research.
Honor of Kings is a 5V5 multiplayer online battle arena (MOBA) game that is extremely difficult and sophisticated and necessitates substantial player cooperation. This makes it the ideal setting for the study and development of AI.
The AI is smarter than you can think
The ability to train AI to become smarter increases with environment complexity. Games offer the ideal R&D environment because they have clear goals and indicators that are simpler to test and iterate, according to Will Yang, general manager of Tencent AI Lab.
In complicated games like HOK, Yang continued, AI may learn to perceive, evaluate, understand, reason, make judgments, and act in real time like people. This suggests that AI has a larger potential to solve issues in a variety of industries, such as robotics, agriculture, transportation, and energy.
The combined power of Tencent AI Arena has enabled the competition to develop into a platform that connects business stakeholders, academic institutions, and research organisations.