Home About us

Robin Li: A "super capable" application is more important than a "super app" of 1 billion DAUs

New Position 2024/07/04 17:04

"In the era of AI, 'super capable' applications are more important than 'super apps' that only look at DAUs." On July 4, Robin Li, founder, chairman and CEO of Baidu, said at the main forum of industrial development of the 2024 World Artificial Intelligence Conference and High-level Conference on Global Governance of Artificial Intelligence that it is necessary to jump out of the thinking logic of the mobile era and avoid falling into the "super application trap". At the same time, Robin Li pointed out that agents are the easiest AI applications to develop, and "they are also the development direction of AI applications that we are most optimistic about."

In his speech, Robin Li once again talked about open source and closed-source large models, saying that open source large models have value in specific scenarios such as academic research and teaching, but they are not suitable for most application scenarios. "When you are in a fiercely competitive environment and need to make your business more efficient and costless than your peers, then the commercial closed-source model is the best way to play."

NewPosition, Robin Li: "Super capable" applications are more important than "super apps" with 1 billion DAUs

Without applications, open-source closed-source models are worthless

Robin Li pointed out that since the beginning of this year, open source and closed-source large models are a controversial topic, but many people confuse the concept of model open source and code open source, he pointed out that model open source can only get a bunch of parameters, but also in SFT, security alignment, even if you get the corresponding source code, I don't know how much and what proportion of data is used to train these parameters, and it is impossible to achieve high flame for everyone to pick up firewood, "Getting these things does not allow you to stand on the shoulders of giants for iterative development." ”

He bluntly said that under the same parameter scale, the ability of the open source model is not as good as that of closed source, "If open source wants to catch up with closed source, it needs to have a larger parameter scale, which means that the inference cost is higher and the response speed is slower." Many people use the open-source model to make a facelift, thinking that it can better serve their own personalized needs, but they don't know that this has become a unique model, which can neither benefit from the continuous upgrade of the basic model, nor share computing power with others. ”

Robin Li admits that the open source model has value in some academic research and teaching fields, and can be used to study the working mechanism of large models and form theories. However, the open source model is not suitable for most application scenarios, and in the fierce business environment, if you want to make business efficiency higher than peers and cost lower than peers, the commercial closed-source model is "the best". He took Baidu's practice in novel creation as an example, when it shifted from the open-source model to the lightweight model, and then to the Wenxin model 4.0, the availability and quality rate of novel generation have been greatly improved, so that online writers are like tigers with wings.

In his view, the focus of the large model is still "volume application", "without application, the basic model alone is worthless, whether it is open source or closed source." ”

A "super capable" app is more important than a "super app".

For AI applications, Robin Li pointed out that in the AI era, "super capable" applications are more important than "super applications" that only look at DAUs, "We must avoid falling into the 'super application trap', and feel that there must be a 1 billion DAU APP to be called successful, which is the thinking logic of the mobile era." He believes that as long as it can produce great gains to industries and application scenarios, the overall value is greater than that of the mobile Internet.

He took the express delivery industry as an example, using the large model ability to process orders, and the express company achieved "one picture, one sentence to send express delivery", no longer need other cumbersome processes, and the time was shortened from more than 3 minutes to 19 seconds. "Moreover, more than 90% of after-sales problems are also solved by large models, and the efficiency improvement is very significant."

In more general areas such as code generation, Robin Li said that Baidu's Wenxin Quick Code has gradually penetrated in various fields, and within Baidu, about 30% of the code has been generated by AI, and the code adoption rate has exceeded 44%.

"Applications are actually not far away," Robin Li said, adding that applications based on basic models have begun to penetrate all walks of life and fields. Two months ago, Baidu announced that the daily call volume of the Wenxin model exceeded 200 million, and recently, the call volume has exceeded 500 million. "In just two months, the number of calls has changed so much, which shows that it represents the real demand behind it, and that someone is using it, and someone is really benefiting from the large model and getting value."

Agents are the most promising direction for AI applications

In terms of the development direction of AI applications, Robin Li is most optimistic about intelligent twins. He said that with the increasing power of the basic model, the development of applications is becoming more and more simple, among them, the simplest is the agent, as long as the workflow is explained clearly in "human words", coupled with a proprietary knowledge base, a very valuable agent can be made, "than the Internet era to make a web page is simpler".

He believes that in the future, various industries such as healthcare, education, finance, manufacturing, transportation, and agriculture will make various agents based on their own scenarios and unique experiences, rules, data, etc., and millions of agents will appear to form a huge ecosystem.

"Search is the biggest entrance for agent distribution," Robin Li said, for example, after the college entrance examination, there are a large number of candidates to fill in the volunteers, need to choose schools and majors, will encounter various problems, at this time, Baidu's college entrance examination agent can answer various questions of candidates, "During the peak period, Baidu's college entrance examination agent has to answer more than 2 million candidate questions every day, a total of only 10 million candidates, indicating that a large proportion of people are using this agent in a day." ”

At the end of the speech, Robin Li once again talked about the classic topic of whether AI will replace human work, he said that AI is currently playing the role of Copilot to assist people in their work, rather than replacing people, and has now generated some new job opportunities, such as data annotators, prompt word engineers, etc. He emphasized that AI will always be just a tool, not a competitor to humans, "We build and apply AI technology to meet human needs, enhance human capabilities, and make human life better." ”

This article is from Xinzhi self-media and does not represent the views and positions of Business Xinzhi.If there is any suspicion of infringement, please contact the administrator of the Business News Platform.Contact: system@shangyexinzhi.com