Tencent to launch AI foundation model for ChatGPT-like applications amid Chinese Big Tech jostling

Tencent executive Li Qiang said the Hunyuan foundation model will be ‘formally released’ soon, with an event to be held on Thursday
China’s first batch of approvals for generative AI services last month has spurred a new round of release announcements for chatbots and related services

Artificial intelligence

Ann Caoin Shanghai

Published: 10:00pm, 6 Sep 2023

Why you can trust SCMP

Chinese social media and video gaming giant Tencent Holdings will soon launch its artificial intelligence (AI) foundation model Hunyuan, which is expected at an event scheduled for Thursday, as the Chinese tech giant jostles for supremacy in the crowded domestic market for large language models (LLMs).

The company’s annual Global Digital Ecosystem Summit takes place on Thursday and Friday in Shenzhen, where there will be a speech about the Hunyuan model, according to the agenda.

The new model, which was under internal testing for a few months, will be “formally released” soon, Li Qiang, a vice-president at Tencent who heads its Government and Enterprise Business division, said at an industry conference held in the southwestern city of Chongqing on Monday, according to a report from Chinese media The Beijing News.

Baidu teases next AI model after chatbot wins Beijing’s approval

The move came after Beijing granted its first batch of approvals for generative AI services in August, giving technology firms including search giant Baidu and AI specialist SenseTime the green light to offer ChatGPT-like chatbots to the public.

That led to a new round of announcements for LLM releases in China, where US-developed services like OpenAI’s ChatGPT and Google’s Bard are officially unavailable.

Baidu, the first major Chinese tech company to launch a ChatGPT-style AI chatbot in March, said opening Ernie Bot to hundreds of millions of internet users would help improve the firm’s foundation model. Those two companies – along with state-backed Zhipu AI and Baichuan, a venture from Sogou co-founder Wang Xiaochuan that is backed by Tencent – issued separate statements last week announcing the end of the “invite-only” and “beta testing” labels attached to their chatbots.

Tencent has yet to release more details about its upcoming foundation model, such as whether it will be available to the general public.

All of Tencent’s internal products will be embedded with or “transformed” by the new model, Li said. The firm built a large-scale computing centre this year with the “latest” graphics processing units, according to Li, who did not specify which chips were installed.

The Shenzhen-based tech giant is among several Chinese Big Tech firms betting big on the industrial application of LLMs. Baidu and Alibaba Group Holding have made similar moves, as the industrial use faces lower regulatory scrutiny and can be commercialised more quickly. Alibaba owns the South China Morning Post.

Baidu co-founder and CEO Robin Li Yanhong said at a company event in Beijing on Tuesday that it was more “meaningful” for entrepreneurs to join the competition of LLM-powered applications rather than building foundation models. “The models themselves do not directly generate value, while the applications based on them are the meaning for the models’ existence,” he said.

Tencent said in June that the company had launched over 50 LLM-enabled industrial solutions covering more than 10 industries, including finance, media, travel and education.

“We will continue to drive innovation, including through generative AI, where we are providing a library of models to our partners via our Tencent Cloud model-as-a-service offering, as well as refining our proprietary foundation model,” Pony Ma Huateng, founder and CEO of Tencent, said during the company’s latest earnings call last month.

Post