miHoYo co-authors science paper outlining bold AI intentions

The paper examines the rise and potential of AI in game development

miHoYo co-authors science paper outlining bold AI intentions

Mobile gaming giant miHoYo has co-authored a paper examining the state of large language models (LLM) and AI in game development, collating 264 different papers on the potential of the technology in game development.

“For a long time, humanity has pursued artificial intelligence (AI) equivalent to or surpassing human level, with AI agents considered as a promising vehicle of this pursuit,” opens the paper. “AI agents are artificial entities that sense their environment, make decisions, and take actions.

In its comprehensive survey on large language model-based applications, the company identifies some must-read papers on the topic, specifically examining “the general conceptual framework for LLM-based agents: comprising three main components: brain, perception, and action, and the framework can be tailored to suit different applications.”

Room to grow

With miHoYo creating some of the most expansive mobile games of all time, including the likes of Honkai: Star Rail and Genshin Impact, it may be hard to believe that they haven’t already reached the peak of what’s possible - however, it seems that the company still believes there’s plenty of “up” to go, with AI as a key aspect of possible expansion.

On X (formerly Twitter), Stanford University senior AI scientist Dr. Jim Fan noted that “If Stanford Smallville ("Generative Agent") can be scaled up to Genshin's or other big MMORPG's player base, that would be quite a spectacle to watch. Both humans and AI agents would have extraordinary emergent behaviors that I couldn't possibly predict.

“Unfortunately, efficiency and cost are key challenges that stand in the way of such mass deployment. For now.”

AI has come a long way in just a few short years. What was considered revolutionary a decade ago, such as Alien Isolation’s use of dynamic AI, is now appearing more and more commonly, but these early steps have opened the door for further development.

The paper identifies the three primary components of LLM - the brain, perception, and action, and how these components can work together to help create a more immersive experience for the player. For example, the brain can analyse data sets and take appropriate action, such as examine current in-game weather conditions and weather reports to identify whether an umbrella is needed in-game.

We listed miHoYo as one of the top 50 mobile game makers of 2023.


Staff Writer

Lewis Rees is a journalist, author, and escape room enthusiast based in South Wales. He got his degree in Film and Video from the University of Glamorgan. He's been a gamer all his life.