In a recent exchange with employees, Robin Lee, the founder, chairman and CEO of Baidu, talked about three major cognitive misunderstandings about the big model in the industry, including big model competition, open source model efficiency, and the development trend of intelligent agents. Robin Lee believes that the gap between big models in the future may become larger and larger. He further explained that the ceiling of the large model is very high, and it is still far from the ideal situation, so the model needs to be constantly iterated, updated, and upgraded quickly; It takes several years or even decades of continuous investment to meet user needs, reduce costs, and increase efficiency.
Rankings do not represent the strength of large models
Robin Lee said frankly that "every time a new model is released, I have to compare it with GPT-4o, saying that my score is almost the same as it, and even exceeds it in some individual items, but this does not mean that there is no gap with the most advanced models."
He explained that many models, in order to prove themselves, will go to the leaderboard after release and guess test questions and answer skills. From the rankings, perhaps the model's capabilities are already very close, but in practical applications, there is still a significant gap.
Robin Lee believes that the gap between models is multidimensional. Industries often focus more on the gaps in understanding, generation, logic, memory, and other abilities, but overlook dimensions such as cost and reasoning speed. Some models can achieve the same effect, but they have high costs and slow inference speed.
In his internal speech, Robin Lee said that the real measure of the capability of the big model should be to see whether it can meet user needs and generate value gains in specific application scenarios, which is the most important.
We should objectively view the efficiency issues of open source models
Robin Lee further explained the misunderstanding of the outside world on the open source model. Before the era of big models, people were accustomed to open source meaning free and low cost, "he explained. For example, open source Linux is free to use because computers already exist. But these are not valid in the era of big models. Big model inference is expensive, and open-source models do not provide computing power. You have to buy your own equipment, which cannot achieve efficient utilization of computing power.
Open source models are not effective in terms of efficiency, "he said." Closed source models should be accurately called commercial models, which are the machine resources and GPUs used by countless users to share research and development costs and inference. GPU usage efficiency is the highest, with Baidu Wenxin Big Model 3.5 and 4.0 having GPU usage rates of over 90%
Robin Lee believes that the open source model is valuable in the fields of teaching and scientific research; But in the business world, when users pursue efficiency, effectiveness, and the lowest cost, open source models lose their advantages.
Intelligent agents are not yet industry consensus
Robin Lee also talked about the development stage of the application of the big model. He believed that Copilot was the first one to assist people; Next is the Agent intelligent agent, which has a certain degree of autonomy and can use tools, reflect, and evolve on its own; If this level of automation continues to develop, it will become an AI worker capable of independently completing various tasks.
At present, agents have attracted more and more attention from large model companies and customers. Robin Lee believes that although many people are optimistic about the development direction of agents, so far, agents have not reached a consensus. There are not many companies like Baidu that consider intelligent agents as the most important strategy and development direction for big models.
Why emphasize intelligent agents? Robin Lee also gave the answer. The threshold for intelligent agents is indeed very low, and many people do not know how to turn large models into applications. However, intelligent agents are a very direct, efficient, and simple way to build intelligent agents on top of models, which is quite convenient. At present, tens of thousands of new intelligent agents are created on the Baidu Wenxin intelligent agent platform every week, and the daily distribution frequency of intelligent agents has rapidly increased to 8 million times.