On April 16, Robin Lee, founder, chairman and CEO of Baidu, shared the specific ideas and tools Baidu has put into practice in the past year to develop native AI applications at the Create 2024 Baidu AI Developers Conference.
Robin Lee said that the big language model itself does not directly create value. Only AI applications developed based on the big model can meet the real market demand. According to their sharing, the specific ideas and tools for developing AI native applications based on large models mainly lie in three points:
The first is MoE. Most large-scale AI native applications in the future will be based on MoE, which is not a general academic concept, but a mixture of large and small models that do not rely on a single model to solve all problems.
The second is the small model. Small models have low inference costs and fast response times. In some specific scenarios, after fine-tuning, their performance can be comparable to that of large models.
The third is intelligent agents. Intelligent agents are currently a very hot topic, and with the improvement of their capabilities, a large number of AI native applications will continue to emerge. The mechanism of intelligent agents, including understanding, planning, reflection, and evolution, enables machines to think and act like humans, autonomously complete complex tasks, continuously learn, achieve self iteration, and self evolution in the environment. (Li Zijian)