US Media: "China's Baidu to Launch Next-Generation AI Model in Second Half of the Year"

Multimodal Capabilities Expected to Improve

China's largest portal search company Baidu is expected to launch a next-generation artificial intelligence (AI) model in the second half of this year.

According to a report by US CNBC on the 12th citing informed sources, the source did not specify the functions of the next-generation AI model 'Ernie 5.0' but said that its multimodal capabilities would be significantly enhanced.

Multimodal AI refers to AI systems that can understand and generate information from various types of data such as text, images, video, and audio.

This report has attracted attention as Chinese companies, including the startup DeepSeek, are competing in AI technology with US companies such as OpenAI, the developer of ChatGPT.

DeepSeek, headquartered in Hangzhou, China, released a low-cost, high-performance AI model last month. Baidu did not respond to CNBC's request for comment. Robin Li (Li Yanhong), CEO of Baidu, said at the World Government Summit held in Dubai this week, "The inference cost of foundation models can basically be reduced by more than 90% over 12 months," adding, "If costs can be reduced by a certain percentage, it means productivity increases by the same amount."

Foundation models can perform various tasks such as understanding language, generating text and images, and natural language communication, and Ernie 5.0 is also reported to be called a foundation model.

Earlier, Baidu was the first Chinese technology company to launch a chatbot similar to ChatGPT called 'Ernie Bot' in March 2023. However, competition has intensified as not only large Chinese tech companies such as Alibaba and ByteDance, the parent company of TikTok, but also startups are rushing to release AI chatbots. Baidu unveiled Ernie 4.0 in October 2023 and released the latest 'Turbo' version last August.