The editor of Downcodes learned that Baidu Chief Technology Officer Wang Haifeng shared the latest progress of Wen Xinyiyan at the "Baidu World 2024" conference, announcing that its user base has exceeded 430 million. This sharing covers many aspects such as the technological upgrade, application expansion and industrial empowerment of the Wenxin large model, demonstrating Baidu's remarkable achievements in the field of artificial intelligence. Wang Haifeng introduced in detail the technological breakthroughs of Wenxin Big Model in knowledge enhancement, retrieval enhancement and intelligent agents, and focused on its application cases in areas such as code generation, image generation and AI talent training.
Baidu Chief Technology Officer Wang Haifeng announced at the "Baidu World 2024" conference that Wen Xinyiyan's user base has reached 430 million. Since its launch on March 16 last year, Baidu Wenxin Big Model has undergone continuous upgrades, and its knowledge enhancement, retrieval enhancement, and intelligent agent technologies have continued to evolve. Wang Haifeng introduced in detail the latest technological progress and applications of Wenxin large models, including retrieval-enhanced Wenxin graph technology, code-free products Miaida, Wenxin fast code, etc.
In terms of retrieval enhancement, Baidu has developed "understanding-retrieval-generation" collaborative optimization technology to improve large model technology and application effects. Vincentian graph technology can generate precise images and can also generate generalized images. Baidu also released the intelligent agent mechanism and developed System 2, the core of which is a thinking model, including understanding, planning, reflection and evolution, which can reliably execute, self-evolve, and white-box the thinking process to a certain extent.
Wang Haifeng also shared the application of multi-agent in coding scenarios - Miaida and Wenxin Quick Code. Miaida adopts the centralized model of intelligent agents to improve the quality and stability of complex code generation. Wenxin Kuaicode upgraded to version 3.0, demonstrating its ability to complete system function development, code testing and problem repair. In the entire development process, multiple intelligent agent pipelines collaborate, which greatly improves the programmer's work quality and efficiency.
Baidu has created a Wenxin large model matrix to meet industry needs, including flagship large models such as ERNIE4.0 Turbo, lightweight models such as ERNIE Speed, and a series of thinking models and scenario models based on basic models to meet the needs of different applications. The Wenxin large model and Feipiao deep learning platform are jointly optimized to improve model effect, training efficiency and inference performance. Wang Haifeng revealed that the number of Feipiao Wenxin developers has reached 18.08 million, serving 430,000 companies and creating 1.01 million models.
Baidu also announced the latest data on artificial intelligence talent training, and has trained 5.92 million AI talents for society. Wang Haifeng revealed that the Wenxin model is still undergoing continuous training, and a more powerful new version is worth looking forward to.
All in all, the rapid development and widespread application of Baidu Wenxin’s large model marks that China’s artificial intelligence technology has made significant progress, and its future development is worth looking forward to! The editor of Downcodes will continue to pay attention to its subsequent progress.