AI News

Li Yanhong Reflects on the Pitfalls Baidu Has Stumbled Into Over the Past 3 Years: Departments Trained Models Separately, Chased the Agent Hype, Now Demands Focus on Efficiency

The PaperUpdated Nov 12 13:56

AI summary

At the opening ceremony of Baidu World 2026, Li Yanhong reflected on the pitfalls Baidu has stumbled into over the past 3 years: in the early days, business departments each trained their own models, which fell behind as soon as the foundation model was upgraded, so it was later changed to unified training by a dedicated model department; during the agent boom, it once again "did whatever was hot," leading to scattered resources, and ultimately it required focus and doing less.

Body · AI translation

Departments trained models in a scattered way and chased agent hype: Li Yanhong reflects on the pitfalls Baidu has stepped into over the past 3 years, and now demands focus and efficiency improvement 2026-11-12 12:33 Source: The Paper ∙ Qikejie>

On November 12, at the opening ceremony of Baidu World 2026, Li Yanhong disclosed some of the pitfalls Baidu had stepped into during the AI transformation era. He said that in the early days, business departments trained their own models, and as soon as the foundation model was upgraded, they fell behind; later, this was changed to unified training by a dedicated model department. During the agent boom, there was also a period of "doing whatever was hot," which led to scattered resources, and ultimately the requirement became to focus and do less.

The Paper reporter Hu Yifan; Editor in charge: Huan Yanhong +1 Favorite #Baidu #Li Yanhong View more

Source

The Paper