"ERNIE Bot has more than 100 million users." On December 28, Wang Haifeng, chief technology officer of Baidu and director of the National Engineering Research Center for Deep Learning Technology and Applications, announced at the 10th WAVE SUMMIT Deep Learning Developers Conference. At the meeting, Wang Haifeng made a keynote speech on the theme of Wenxin Plus PaddlePaddle, Flying to the Milky Way, and shared the latest achievements of PaddlePaddle and Wenxin.
PaddlePaddle developers have reached 10.7 million
The WAVE SUMMIT Deep Learning Developer Conference began in April 2019 and meets developers twice a year. It has now been five years and ten sessions.
Looking back on five years, the conference has witnessed Baidu's forward-looking judgment on artificial intelligence technology and industry trends, guiding the direction of technological innovation and industrial practice. In 2019, Wang Haifeng proposed at the first conference that deep learning framework is the operating system of the intelligent era. The universality of deep learning, as well as the development of deep learning frameworks and platforms, promote the standardization, automation, and modularization of artificial intelligence, and enter the stage of industrial large-scale production. In 2020, Wang Haifeng proposed to build a new type of AI infrastructure, accelerate industrial intelligence through cloud intelligence integration, upgrade AI large-scale production platforms to cloud intelligence integrated new infrastructure, and lay a solid foundation for industrial intelligence. In 2021, Wang Haifeng stated that artificial intelligence exhibits the characteristics of "integrated innovation" and "lowering threshold": on the one hand, there is an increasing number of integrated innovation in AI technology and industry; On the other hand, although AI technology is becoming increasingly complex, the threshold for AI development and application is getting lower and lower. In 2022, Wang Haifeng further proposed that the addition of large models to deep learning platforms would connect the entire AI industry chain from hardware adaptation, model training, inference deployment, to scenario applications, solidifying the foundation of industrial intelligence. This year, the emergence of big language models has brought dawn to general artificial intelligence.
Over the past five years, in the process of continuous technological innovation and the development of the enabling industry, PaddlePaddle itself has also been constantly upgrading, from the deep learning framework to the platform ecology, to develop into an industrial level open source platform for deep learning with advanced technology and rich functions. PaddlePaddle integrates core framework, basic model library, development kit, tool components, and the Galaxy community that helps developers grow. It has key core technologies such as dynamic and static unified deep learning framework, end-to-end adaptive large-scale distributed training, cloud side full scene high-performance reasoning, etc.
The ecology of PaddlePaddle is becoming more and more prosperous. In 2019, the number of developers on the PaddlePaddle platform was 1.5 million. By August this year, the Wave Summit had reached 8 million. The number of enterprises served and the number of models created based on PaddlePaddle also grew rapidly. Wang Haifeng announced the latest achievements of PaddlePaddle ecology on the spot. By the end of December 2023, PaddlePaddle had gathered 10.7 million developers, served 235000 enterprises and institutions, and created 860000 models based on PaddlePaddle.
The number of ERNIE Bot users has exceeded 100 million, and the daily number of questions has grown rapidly
It is understood that Baidu has been deeply involved in the research and development of pre training models since 2019, and has released the Wenxin Big Model 1.0. After nearly four years of accumulation, Baidu took the lead in releasing ERNIE Bot, a big language model for knowledge enhancement, among the world's major technology plants in March this year. In October, ERNIE Bot's basic model was upgraded to 4.0, and the four basic AI capabilities of understanding, generation, logic and memory were comprehensively improved. The overall performance of Wenxin Big Model 4.0 has increased by 32% in the past two months.
Wang Haifeng disclosed on the spot that the number of ERNIE Bot users has exceeded 100 million. Since the opening of ERNIE Bot to provide services to the public on August 31, the number of users' questions has risen all the way, basically in step with the improvement of the effect of Wenxin's big model. More and more users trust and use ERNIE Bot.
Wang Haifeng finally said: "In the five years and ten sessions, we will work hard with all the developers, and we are willing to continue to work together with all the developers, and with the support of PaddlePaddle and Wenxin, we will go to the star sea of general artificial intelligence!"