Baidu's ERNIE Bot Surpasses 100 Million Users, PaddlePaddle Developers Reach 10.7 Million
-
"ERNIE Bot's user base has exceeded 100 million." On December 28, Wang Haifeng, CTO of Baidu and Director of the National Engineering Research Center for Deep Learning Technology and Applications, announced this at the 10th WAVE SUMMIT Deep Learning Developer Conference. During the event, Wang Haifeng delivered a keynote speech titled "ERNIE and PaddlePaddle: Soaring to the Stars," sharing the latest achievements of PaddlePaddle and ERNIE.
PaddlePaddle developers have reached 10.7 million
The WAVE SUMMIT Deep Learning Developer Conference began in April 2019 and has been held twice a year, now marking its fifth year and tenth edition.
Looking back over five years, the conference has consistently witnessed Baidu's forward-looking judgments on artificial intelligence technology and industry trends, guiding the direction of technological innovation and industrial practices. In 2019, Wang Haifeng proposed at the inaugural conference that deep learning frameworks are the operating systems of the intelligent era. The general characteristics of deep learning, along with the development of deep learning frameworks and platforms, have propelled AI into standardization, automation, and modularization, entering the stage of industrial-scale production. In 2020, Wang Haifeng suggested building a new type of AI infrastructure, integrating cloud and intelligence to accelerate industrial intelligence, upgrading the AI production platform into a new infrastructure that combines cloud and intelligence, laying a solid foundation for industrial intelligence. In 2021, Wang Haifeng noted that AI is exhibiting characteristics of "convergence innovation" and "lowering barriers": on one hand, there is increasing convergence innovation in AI technology and industry; on the other hand, despite the growing complexity of AI technology, the barriers to AI development and application are becoming lower. In 2022, Wang Haifeng further proposed that deep learning platforms combined with large models bridge the entire AI industrial chain from hardware adaptation, model training, and inference deployment to scenario applications, solidifying the foundation for industrial intelligence. This year, the emergence of large language models has brought a glimmer of hope for general artificial intelligence.
Over the past five years, in the continuous process of technological innovation and empowering industrial development, PaddlePaddle itself has been constantly upgrading, evolving from a deep learning framework to a platform ecosystem, developing into a technologically advanced, feature-rich industrial-grade deep learning open-source platform. PaddlePaddle integrates core frameworks, basic model libraries, development kits, tool components, and the Galaxy community that supports developer growth, featuring key core technologies such as a unified dynamic and static deep learning framework, end-to-end adaptive large-scale distributed training, and high-performance inference for cloud-edge-end full scenarios.
The PaddlePaddle ecosystem continues to flourish. In 2019, the platform had 1.5 million developers, which grew to 8 million by the Wave Summit in August this year. The number of enterprises served and models created based on PaddlePaddle have also experienced rapid growth. Wang Haifeng announced the latest achievements of the PaddlePaddle ecosystem: as of December 2023, PaddlePaddle has gathered 10.7 million developers, served 235,000 enterprises and institutions, and facilitated the creation of 860,000 models.
ERNIE Bot's user base has surpassed 100 million, with daily queries growing rapidly.
It is reported that Baidu has been deeply involved in pre-training model research and development since 2019, releasing ERNIE 1.0. After nearly four years of accumulation, Baidu took the lead in launching the knowledge-enhanced large language model ERNIE Bot in March this year. In October, the foundation model of ERNIE Bot was upgraded to version 4.0, with comprehensive improvements in the four fundamental AI capabilities: understanding, generation, logic, and memory. Over the past two months, the overall performance of ERNIE 4.0 has improved by another 32%.
Wang Haifeng disclosed on-site that Ernie Bot's user base has surpassed 100 million. Since its public service launch on August 31, user queries have consistently increased, closely aligning with the performance improvements of the Ernie large model. More and more users are trusting and utilizing Ernie Bot.