Baidu founder Robin Li: We are about to enter an AI-native era

"Finance" new media writer Wang Jingya/text Gao Suying/editor

"We are about to enter an era of native AI, an era where humans and machines interact through prompts, and the future will be generated by us together.". On October 17, Baidu founder, chairman and CEO Robin Li said at the 2023 Baidu World Conference.

He announced on the spot that Baidu Wenxin was officially upgraded to version 4.0. Compared with the previous version, the new version has achieved significant improvements in the four major capabilities of understanding, generation, logic and memory, and its overall level is not inferior to GPT4. This is currently Baidu's strongest Wenxin large model, which has achieved a comprehensive upgrade of the basic model.

Li Yanhong demonstrated the characteristics and application scenarios of Wen Xinyiyan’s four abilities of understanding, generation, logic, and memory. Robin Li believes that these capabilities are not available in the past era, so they can open up unlimited space for innovation.

Specifically, in terms of understanding ability, AI has developed from an "artificial retard" that cannot understand human speech to one that can understand almost all speech, and even understands what the user is saying better than the user's friends and colleagues. In terms of generation capabilities, based on a picture material and several key words provided by Robin Li, Wen Xinyiyan generated 1 advertising video, 5 pieces of advertising copy and 1 poster in just 3 minutes. Based on this ability, Baidu has launched Qingduo, an AIGC marketing creative platform.

In terms of logical ability, the application of the Wenxin large model is particularly obvious in scenarios such as solving mathematical problems and summarizing knowledge points. Robin Li said that in addition to problem solving, logical capabilities are required for route planning on smart maps, complex tasks handled by smart assistants, traffic light control in smart transportation systems, etc. Robin Li pointed out that in terms of memory ability, whether the AI ​​remembers what the user said and whether the content generated by the AI ​​is inconsistent before and after is an important indicator to distinguish the intelligence of a large model. Multiple rounds of dialogue are the embodiment of memory ability.

It should not be ignored that the four major capabilities of the large model do not exist independently, but are complementary to each other in specific scenarios. In Robin Li's view, understanding, generation, logic, and memory capabilities are the basis for the survival of all AI native applications. For example, when creating advertising copy, you need to understand the creative theme, clarify the creative logic, and maintain consistency through memory. In solving problems, these four abilities also need to be comprehensively applied.

It is worth mentioning that the ultimate goal of large-scale model technology from all walks of life is still to serve people, and practical application is the key to the development of AI. "AI native applications are applications developed based on the understanding, generation, logic and memory capabilities of large models." Robin Li believes that without rich AI native applications built on the basic model, the basic model has no value.

Robin Li demonstrated more than 10 AI native application cases based on Wen Xinyiyan's reconstruction of Baidu Search, Ruliu, Maps, Netdisk, and Wenku, hoping to inspire developers to work together to make more amazing things. AI native applications. In his view, "China has rich application scenarios, and Chinese users are willing to embrace new technologies. With advanced basic large models, we can build a prosperous AI ecosystem and jointly create a new round of economic growth."

When developing AI native applications, the basic capabilities of large models are crucial. Robin Li said that API is the main way for AI native applications to call basic large models. Currently, there are 42 mainstream large models settled on the Qianfan Large Model Platform, covering nearly 500 scenes in various industries.

It is worth noting that large model reconstruction will not only affect online applications, but will also affect offline work and life. A large number of AI native applications will continue to emerge, promoting the deep integration of digital technology and the real economy. At present, large model technology has been applied in manufacturing, energy, electric power, chemical industry, transportation and other real industries, and is becoming an important driving force for new industrialization.

Robin Li believes that a new world and a new future will be generated through prompts from every enterprise, every developer, and every user. Future AI native applications must be multi-modal and will reconstruct the physical world in addition to the information world.

The above is the detailed content of Baidu founder Robin Li: We are about to enter an AI-native era. For more information, please follow other related articles on the PHP Chinese website!