The logo of Baidu’s AI chatbot Ernie Bot is displayed near a screen showing the Baidu logo, June 28, 2023. Reuters-Yonhap
Chinese technology giant Baidu is embracing DeepSeek’s artificial intelligence (AI) models by integrating it with its search engine, fast on the heels of Tencent Holdings, which recently adopted the start-up’s tech into Weixin, the country’s largest social media platform.
Baidu will fully connect both DeepSeek and its own Ernie large language models (LLMs) with its search engine to offer a “more diversified search experience,” the company announced on Sunday.
The search giant will also add DeepSeek to its LLM platform for developers, Baidu said in a separate statement.
Baidu is embracing DeepSeek just after its rival Tencent announced beta testing of the AI models on its super app Weixin, as it calls WeChat in mainland China. Some users were given access to the DeepSeek-R1 model through Weixin’s search bar over the weekend, potentially widening the exposure of the popular Chinese AI models to the platform’s 1.3 billion active users. It is currently only being offered to mainland China accounts.
Baidu’s share price plunged around 8 percent in afternoon trading in Hong Kong, marking its sharpest fall since last November. The company is slated to report its fourth quarter earnings on Tuesday.
Hangzhou-based DeepSeek has become the hottest AI start-up in China after making waves in the global tech community. News of two advanced models, V3 and R1, that were trained at a fraction of the cost of other leading models rocked Wall Street last month, sending AI-related stocks plummeting.
LLMs are the technology underpinning generative AI services like ChatGPT, Baidu’s Ernie Bot and DeepSeek’s namesake chatbot.
A key to DeepSeek’s explosive success is that its models are open source, allowing anyone to use and modify them. This has spurred competitors to adopt their own open-source strategies, including Baidu, which has largely focused on closed-source AI development.
Last week, Baidu said the next generation of its Ernie LLM coming on June 30 will be open source. It also announced that its Ernie Bot service will be free to use from April 1.
The move also came as Chinese companies and local government have rushed to deploy DeepSeek’s products on their own systems.
The cities of Guangzhou and Shenzhen, in southern Guangdong province, said on Sunday that they have integrated the DeepSeek-R1 reasoning model in their civil service platforms.
The municipal government of Guangzhou, capital of Guangdong, said it has adopted DeepSeek-R1 and other LLMs for domestic hardware and would use them to enhance various public services such as interpreting policy and task dispatching for the government hotline.
Shenzhen’s Longgang district adopted the R1 model on February 8, making it the first district government in the country to widely adopt the start-up’s low-cost, high-performance model.
Read the full story at SCMP.
