On April 16, at the Baidu AI developer conference, Robin Lee, the founder, chairman and CEO of Baidu, disclosed the progress of ERNIE Bot in the year of its release. The number of users has exceeded 200 million, the average daily API calls have exceeded 200 million, and the number of customers or enterprises served has reached 85000. After a year of development, the reasoning cost of the Wenxin model has decreased to 1% of its original level. 27% of the new code added by Baidu every day is automatically generated by Comate, an intelligent code assistant based on the Wenxin model.
Robin Lee also said that the future large AI native applications are basically MoE (mixed expert model), which means that there will be different sizes of mixed applications, which should be adapted according to specific scenarios. In some specific scenarios, the efficiency of using fine tuned small models can achieve the effect of large models. Not long ago, Robin Lee said in his internal speech that the strongest basic models in both China and the United States are closed source, and all kinds of small models and the best small models are distilled from large models. The model created by reducing dimensions through large models is better.