Robin Li said that in today's large model ecosystem, open source models will become increasingly backward.

On April 16, CNMO learned that Baidu founder and CEO Robin Li declared at the Create 2024 Baidu AI Developer Conference that in today’s large model ecosystem, open source models will become increasingly lagging behind. He explained that Baidu's basic model Wenxin 4.0 can be flexibly tailored according to different needs in terms of effects, response speed, and inference cost to generate streamlined models suitable for various scenarios, and supports fine-tuning and post pretraining. Compared with directly using open source models, the models cut out by Wenxin 4.0 perform better at the same size and cost less with the same effect. Therefore, he predicts that open source models will fall further and further behind.

In the previous Baidu financial report conference call, Robin Li revealed that Baidu Intelligent Cloud’s revenue in the fourth quarter reached 8.4 billion yuan, of which large The model contributed approximately 660 million yuan in incremental revenue to the cloud business. At present, the daily calls of Wenxin large models have exceeded 50 million times, with a quarter-on-quarter growth of up to 190%. In December last year, the number of enterprises calling Wenxin Big Model reached 26,000, a quarter-on-quarter increase of 150%. Samsung, Honor, Autohome and other companies have become its partners.

Since its release, Baidu has continued to reduce the inference cost of the Wenxin large model, and has successfully reduced it to 1% of the version in March last year, further improving its market competitiveness.

Li Yanhong said that in the future, multi-modality or multi-modal fusion, such as text to video, is a very important direction for basic model development and is also the direction of AGI (General Artificial Intelligence). a necessary direction for intelligence). Baidu has already invested in these areas and will continue to invest in the future.

The above is the detailed content of Robin Li said that in today's large model ecosystem, open source models will become increasingly backward.. For more information, please follow other related articles on the PHP Chinese website!