科大讯飞作为中国领先的智能语音和人工智能公司，在人工智能大模型领域持续发力。公司副总裁、研究院院长刘聪表示，讯飞研究院致力于将技术研发与业务需求紧密结合，不追求所有功能都领先，而是聚焦于满足特定业务场景的需要。
自2005年成立以来，讯飞研究院一直是科大讯飞技术发展的核心力量。在当前的大模型浪潮中，讯飞研究院不仅在迭代自己的基座模型，还深入研究业务需求，推动人工智能大模型技术在教育、医疗、汽车、办公等多个领域的应用。
2022年12月，讯飞启动了“1+N人工智能大模型技术及应用”专项攻关，“1”代表通用人工智能大模型底座，“N”代表将人工智能大模型技术应用在不同行业领域。经过一年的发展，讯飞已经将星火大模型迭代至4.0版本，并逐步提升了模型的性能，对标国际领先的模型。
刘聪透露，今年的研发重点将更加注重大模型与小模型的并行发展，以及深入解决系统化问题。他认为，选择大模型还是小模型取决于具体场景的需求。在某些领域，如数学能力、语音交互等，讯飞研究院将继续加大研发力度，而像文生图、虚拟人等业务关联度较低的领域，则将重点放在资源投入的有效性上。
在应用方面，讯飞研究院的研发工作与业务部门紧密合作，确保技术的落地能够为用户带来实际价值。刘聪强调，讯飞之所以能将技术成功转化为产品，关键在于对行业场景和技术积累的深入理解，以及对技术成熟度的准确把握。
未来，科大讯飞将继续在大模型领域深耕，通过技术与业务的深度融合，推动人工智能技术的实际应用，为用户创造更大的价值。
英语如下：
Title: iFLYTEK Focuses on Business Needs to Develop a Focused, High-End AI Model Capability
Keywords: iFLYTEK, Model Capability, Business Needs
Content:
As a leading Chinese company in intelligent voice and artificial intelligence, iFLYTEK has been actively involved in the field of artificial intelligence models. According to Liu Cong, Vice President and President of the Research Institute, the Research Institute is committed to closely integrating technology development with business needs. Instead of chasing every cutting-edge feature, it focuses on meeting the specific needs of certain business scenarios.
Since its establishment in 2005, the Research Institute has been the core driving force for iFLYTEK’s technological development. In the current wave of large model research, the Research Institute not only iterates its foundational model but also delves into studying business needs, driving the application of artificial intelligence large model technology in various fields such as education, healthcare, automotive, and office work.
In December 2022, iFLYTEK initiated a special project on “1+N Technology and Application of Artificial Intelligence Large Models,” where “1” represents the universal artificial intelligence large model foundation, and “N” represents the application of AI large model technology in different industries. After a year of development, iFLYTEK has iterated its Spark model to version 4.0 and gradually improved its performance, benchmarking against leading international models.
Liu Cong revealed that this year’s research focus will be more on the parallel development of large and small models, as well as in-depth solutions to systematic issues. He believes that the choice between large and small models depends on the specific needs of the scenario. In areas such as mathematical ability and voice interaction, the Research Institute will continue to increase its research efforts. In contrast, for business-related areas with lower ties, such as text-to-image generation and virtual human, the focus will be on the effectiveness of resource allocation.
In terms of application, the Research Institute’s development work is closely coordinated with business departments to ensure that the technology is successfully implemented to bring real value to users. Liu Cong emphasized that the key to iFLYTEK’s successful conversion of technology into products lies in a deep understanding of industry scenarios and technological accumulation, as well as an accurate grasp of technological maturity.
Looking forward, iFLYTEK will continue to delve into the field of large models, promoting the practical application of artificial intelligence technology through the integration of technology and business, and creating greater value for users.
【来源】https://mp.weixin.qq.com/s/bOOJbcCvKaopP3zNtAfAQA
Views: 4