Chinese Tech Leaders Demand Practical AI Application at World AI Conference

China has too many AI language models and not enough practical applications, Baidu CEO Robin Li said on a panel at the World Artificial Intelligence Conference (WAIC).

The Baidu founder and CEO said the overabundance of Chinese language models has resulted in “intense competition,” according to the South China Morning Post.

Several major technology companies in China rushed to build their own language models in the wake of ChatGPT, as OpenAI’s technology was restricted in the country. Among those companies was Li’s Baidu, which built ERNIE, claiming it is on par with OpenAI’s ChatGPT.

Beyond the underlying model, Baidu’s own Ernie is offered as a chatbot for enterprises via the Qianfan Foundation model platform.

In addition to Baidu's model, Li claims that more than 100 models are competing for a share of the Chinese AI market.

Among the popular Chinese models include Alibaba’s Qwen system, Qingyan from Zhipu and Baichuan from Tencent subsidiary Baichuan Intelligence.

According to Li, the glut of models has seen developers become more focused on outperforming rivals than building actual AI applications to apply their models.

“I’ve noticed that many people still primarily focus on foundational models,” Li said. “But I want to ask: how about real-world applications? Who has benefited from them?”

Related:China Promotes Global AI Cooperation With New Shanghai Declaration

To attract users, the South China Morning Post reports that local vendors, including Alibaba, Tencent and the owner of TikTok, ByteDance, have all reduced the costs of running their models.

Li said at WAIC that the excess of models meant developing wasted “significant resources, particularly computing power.”

Compared to American companies building generative AI systems, Chinese entities are hamstrung by computational limits. The Biden administration blocked exports of powerful chips from vendors like Nvidia to China over national security fears.

Li’s comments in Shanghai repeat those he made at the 2023 X-Lake Forum at Shenzhen University last November, where he said China has “too many foundation models and very few AI native applications.”

His views in Shanghai were echoed by Xu Li, CEO and co-founder of SenseTime. On the same panel, he said that application development was the key to determining whether this era is a critical moment for AI.

Similar to Baidu, SenseTime offers AI services beyond a mere model, including a chatbot called SenseChat and image generator platforms.

At WAIC, the company, which is subject to U.S. sanctions, unveiled Vimi, a controllable AI avatar video generator capable of creating short video clips displayed onto an avatar from still images.

Related:Tesla Optimus Humanoid Robot Draws Crowds at World AI Conference

At the event, Xu said the AI industry has not reached a critical stage “because it has not yet penetrated any applications in any vertical industries that have caused widespread change.”