【作 者】：
卢 宇， 余京蕾， 陈鹏鹤， 余胜泉
【关 键 词 】：
多模态大模型； 人工智能教育应用； 多模态汉字学习； 教育大模型
【栏 目】：
教育数字化
【中文摘要】：
多模态大模型逐渐成为人工智能领域研究的热点，目前已在通用领域有显著进展，但在教育领域仍处于起步阶段。文章提出可以构建教育领域通用大模型，并使其通过下游任务适配形成三类多模态教育大模型，从而形成三种典型教育应用，即教学资源自动生成、人机协同过程支持与教师教学智能辅助。在此基础上，文章以“多模态汉字学习系统”为例，利用多模态大模型实现跨模态释义生成，展示了多模态大模型在辅助语言学习方面的应用潜力。最后，文章针对教育领域通用大模型研究、多模态教育大模型的创新应用及其带来的潜在风险与可能触发的教育变革，提出针对性的建议与展望。
【英文摘要】：
Large multimodal models have gradually become a hot topic of research in artificial intelligence, and have significant progress in the general field. But they are still in the initial stage in the education field. This paper proposes to build a general large model in education and adapt it to three types of educational large multimodal models through downstream tasks, which constitutes three typical applications in education, namely, automatic generation of learning resources, human-AI collaboration, and intelligent teacher teaching assistance. Based on that, this paper takes "multimodal Chinese character learning system" as an example and uses large multimodal model to realize cross-modal interpretation generation, demonstrating the potentials of large multimodal model in assisting language learning. Finally, this paper proposes suggestions and prospects on the research of general large models in education, the innovative applications of educational large multimodal models, and the potential risks and changes in education that may be triggered by them, respectively.