Researchers have developed multiple artificial intelligence (AI) agents that can learn to collaborate to defeat multiple enemies in a real-time strategy computer game, a study released Monday by University College London (UCL) showed.
Single AI agents have been proven to be able to beat the most accomplished human players in the Go and card games. But in this study, researchers demonstrated that multiple AI agents could work together in playing StarCraft, which is considered one of the most difficult games for computers to handle, with far more parameters than Go.
“A major challenge with StarCraft is that the number of agents playing is dynamic, causing the parameters of the model to constantly fluctuate. This forces the AI agents to develop sophisticated behaviors through multi-agent learning, rather than just using a joint learner method,” said Jun Wang from UCL, who led the study.
The discovery provides valuable insight into the way AI agents learn how to communicate and coordinate. This could help scientists develop the next generation of AI for use in large-scale, real-world applications, advancing the AI currently used for gaming on the stock markets, predicting user interests and as competing agents bidding on online advertising exchanges, according to UCL.
The discovery creates “opportunities to develop better models to help understand how language and social structures evolve, or to better predict economic outcomes where each AI agent might represent a company,” Wang said.