Tencent releases AI model for business as competition in China heats up

Chinese tech giant Tencent will launch its “Hunyuan” artificial intelligence model for business use at its annual summit on Thursday, said Dowson Tong, CEO of the smart industry group and cloud at Tencent, told CNBC in an exclusive interview ahead of the event.

The news comes days after Baidu unveiled a series of AI-powered apps on Tuesday following more regulatory support.

Tencent said it is internally testing the Hunyuan AI model in advertising and fintech. The gaming and social media giant is also set to release an AI chatbot on Thursday, the company said in an online post.

Tong told CNBC that Tencent is integrating Hunyuan’s capabilities with existing products for video conferencing and social media.

The company operates WeChat, a widely used messaging and payment app in China, and video conferencing platform Tencent Meet.

Baidu and several other Chinese companies have been given the green light in the past few weeks to release AI-powered chatbots to the public.

Similar to ChatGPT, bots aim to answer queries in a chat-like fashion – but mostly in Chinese. Some, such as Baidu’s Ernie bot, also convert text to images and videos with the help of plugins.

OpenAI’s ChatGPT isn’t officially available in China yet. The chatbot release is subject to China’s new synthetic AI regulation effective August 15.

When asked about regulations, Tong pointed out that such artificial intelligence is so new that no one knows what impact it will have on society.

“It is wise to install some railings in place,” he said. That will help ensure the technology or service provided is of high enough quality, he said, that they don’t create and distribute misinformation.

Chinese authorities said the “temporary” regulations that took effect last month would not apply to companies developing AI technology as long as the product is not made available to the general public.

That’s more relaxed than a draft published in April that said the upcoming regulations would apply even at the research stage.

While Beijing appears to be more supportive of AI than initially feared, Chinese companies also face US restrictions on acquiring advanced semiconductors. The most advanced versions of high-tech chips, called graphics processing units (GPUs), allow companies to train AI models.

“The restrictions we are facing will hinder progress, the pace of development,” Tong told CNBC in response to a question about the US restrictions.

He noted that demand for computing power in general far outstrips supply in China. To mitigate the shortfall, he said companies are “focusing on specific use cases, building models that scale.”

“And we expect that the supply of compute GPUs will be larger in the coming months and so the development of these technologies can be faster.”

Tencent is just one of many companies in China — from startups to phone maker Huawei — that have rushed to announce AI products this year. In August, Alibaba announced it would open its own AI model to third-party developers.

Tencent’s Tong said artificial intelligence requires industry-specific training for the technology to create value. He lists business use cases in travel, finance, public services, and customer service.

“In fact, we believe that many different customers will benefit more by leveraging open source models and using their own enterprise data to train our models,” he said. their own to meet very specific needs in their industrial use cases”.

That designated use can also help protect data, he said.