On April 16, Create 2024 Baidu AI Developer Conference was held in Shenzhen. With the theme of "Creating the Future", the conference focused on the hottest topic of AI. At the conference, Robin Li, founder, chairman and CEO of Baidu, delivered his first speech in 2024, "Everyone is a Developer", to showcase the latest achievements of Baidu's AI technology for developers around the world.
At the Create conference, Robin Li said: "Natural language will become the new general-purpose programming language, as long as you can speak, you can become a developer and change the world with your creativity." And bring three AI development artifacts, AgentBuilder, AppBuilder, and ModelBuilder, making it possible for "everyone is a developer".
01
From large models to large applications, unleash the true value of large models
Generative AI has brought cross-generational development opportunities, and under Robin Li's leadership, Baidu has become the world's first technology giant to release GPT models. From the "pioneer" of the domestic large-scale model platform to the large-scale model "super factory" today, Baidu's pace of innovation has never stopped.
Large-scale model technology has developed explosively in the past year, landing in thousands of industries. In Robin Li's view, Baidu, as a technology company, is positioned to provide everyone with the development tools they need as much as possible, including one basic model series and three AI development tools. These three tools all represent advanced productive forces. "After more than a year on the job, Comate has entered tens of thousands of companies such as Himalaya, Mitsubishi Elevator, iSoftStone, etc., and the code adoption rate has reached 46%, and 27% of Baidu's new code every day has been automatically generated by Comate. ”
According to reports, Wenxin Yiyan was released on March 16 last year, the number of users exceeded 200 million, the average daily API call volume also exceeded 200 million, the number of customers served reached 85,000, and the number of AI native applications developed using the Qianfan platform exceeded 190,000.
Robin Li has repeatedly emphasized in public that large language models do not directly create value, and AI applications developed based on large models can meet real market demand.
02
"In a word", everyone can become a developer by creating an agent
"AI is setting off a creative revolution, and developing applications in the future is as simple as shooting a short video, everyone is a developer, and everyone is a creator. ”
At the conference, Robin Li officially released the tool version of Wenxin Model 4.0. This tool version allows developers to experience the code interpreter function, which can not only process and analyze complex data and files through natural language interaction, but also generate charts or files, which can quickly gain insight into the characteristics of the data, analyze changing trends, and provide support for subsequent decision-making.
It is worth mentioning that compared with a year ago, the algorithm training efficiency of Wenxin's large model has been increased to 5.1 times, the average weekly training efficiency has reached 98.8%, the inference performance has been improved by 105 times, and the cost of inference has been reduced to 1%.
In other words, customers used to make 10,000 calls a day, but now they can make 1 million calls a day at the same cost. Robin Li said that Baidu has a full-stack layout in the four-layer architecture of chips, frameworks, models, and applications, and through end-to-end optimization, it has continuously reduced costs, so that more people can use large models to do AI applications efficiently and cheaply.
In addition, Robin Li shared some of Baidu's specific ideas and tools for developing AI native applications.
One is MoE. In the future, large-scale AI native applications will basically be based on the MoE architecture, that is, a mix of large and small models, which does not rely on one model to solve all problems, and the second is a small model. The inference cost of small model is low, the response speed is fast, and in some specific scenarios, the use effect of small model after SFT fine-tuning can be comparable to that of large model; The agent mechanism includes understanding, planning, reflection, and evolution, which allows machines to think and act like humans, complete complex tasks autonomously, and continuously learn, iterate, and evolve in their environment. In some complex systems, different agents can also interact with each other and cooperate with each other to complete tasks with higher quality.
"These agent capabilities, the ability to reflect, to plan, to plan themselves, we have developed and are fully open to developers. Robin Li announced that the three "out-of-the-box" tools are: agent development tool AgentBuilder, AI native application development tool AppBuilder, and model customization tool ModelBuilder of various sizes.
Based on these directions, Robin Li said that large models and generative AI will revolutionize the developer community. In the past, developers used code to change the world, but in the future, natural language will become the ubiquitous programming language. As long as you can speak, you can become a developer and change the world with your creativity.
03
The basic model + three AI development tools will set off a productivity revolution in the AI era
The revolutionary value brought by large models is not only a technological breakthrough, but also their real value is precisely the potential for application explosion due to versatility. The basic model + three AI development tools bring productivity changes.
Presently. Baidu AI has been applied to search, live broadcasting, e-commerce, smart applets, smart small screens, smart speakers, smart driving and other aspects. Since the launch of the new AI function of Baidu Wenku, the cumulative number of AI users has exceeded 100 million, and the cumulative number of use of the function has exceeded 800 million; Baidu New Search, among the existing Q&A needs of Baidu Search every day, the proportion of the first one has reached 70%, and the average daily new Q&A demand of Baidu Search has exceeded 50 million; Wenxin Yiyan has become a large model product with more than 200 million users.
At the ecological level, we continue to establish cooperative relations with governments, enterprises, developers, schools, social organizations, etc. It not only provides partners with intelligent solutions and intelligent empowerment platforms, but also works with entrepreneurs and developers to promote the prosperity of the large model ecosystem.
"Looking to the future, I believe that multimodal large models, or the integration of text, images, voice, video, etc., is a very important long-term development direction for basic models, and it is the only way to AGI. Baidu has a long-term investment in these areas and will keep up to date with technological progress. Robin Li said.
"Today's China, with 1 billion Internet users, a strong basic model, enough AI application scenarios, and the world's most complete industrial system, the state is also vigorously encouraging and supporting the 'artificial intelligence +' action, and every person and every enterprise only needs to make full use of these tools to unleash unlimited creativity and productivity. ”
Facing the future, Robin Li said, "Everyone can become a developer, and the future will be a future created by developers together!"