• When it comes to gaming, the use of ChatGPT-like generative artificial intelligence is still in an exploratory phase, according to Liang Chen, general manager of Tencent Cloud's internet industry department.
• Chen said that AI has long been used by gaming companies for pattern recognition and other functions, but effective use of generative AI comes with high costs.
• Despite being in an exploratory period, Chen said he expected the business' investment in AI to increase every year.

BEIJING — When it comes to gaming, the use of ChatGPT-like generative artificial intelligence is still in an exploratory phase, according to Liang Chen, general manager of Tencent Cloud's internet industry department.

Parent company Tencent is one of the world's largest gaming companies and owns U.S.-based Riot Games, home of popular titles such as League of Legends and Valorant. Chen said his role is to help game manufacturers improve their productivity and efficiency with technology, including support for the overseas expansion of Tencent's hit mobile game Honor of Kings this year.

Chen said that AI has long been used by gaming companies for pattern recognition and other functions, but effective use of generative AI comes with high costs.

OpenAI's ChatGPT chatbot and similar technologies have enabled AI to better understand and generate content in a human-like way. The tools can also produce inaccurate or illogical content.

"My personal view is that large-scale application [of generative AI] will still take some time," Chen told CNBC in an interview Friday on the sidelines of the annual ChinaJoy gaming conference in Shanghai. That's according to a CNBC translation of his Mandarin-language remarks.

The most pressing issue is "how we can balance costs and results, in order to scale," Chen said.

Chen shared examples such as having to train generative AI to only return historically accurate answers when a user interacts with a virtual character inside a game set several hundred of years ago. For such functions, he said the company is using Tencent's Hunyuan AI model and other in-house developed models.

He declined to comment on Nvidia, whose most advanced chips are prized for training AI models. The U.S. has restricted exports of those semiconductors to China.

OpenAI's ChatGPT and underlying tech for building applications aren't officially available in China. Tencent, Baidu, Alibaba and many startups have created local alternatives.

Tencent's AI Lab in March announced a tool called GiiNEX Game AI Engine that incorporates generative AI for developing city scenes in games.

Chen said that in the last year or so, Tencent has already been using AI tools for 3D rendering, cutting time spent on that step down from at least five days to around two hours.

He said that doesn't necessarily shorten the entire production time as demand for quality goes up and companies need to determine how interested users are in the game.

Tencent's domestic games revenue fell by 2% in the first quarter from a year ago to 34.5 billion yuan ($4.75 billion), while that of international games rose by 3% to 13.6 billion yuan.

Despite being in an exploratory period, Chen said he expected the business' investment in AI to increase every year.

"It is hoped that one day there will be one or several cases that can be replicated quickly, and the [investment] proportion may suddenly increase then," he said.