【作 者】：
蒋贵友， 殷文轩
【关 键 词 】：
大语言模型； 大学教学； 生成式人工智能； 教学生态； 人机融合
【栏 目】：
历史与国际比较
【中文摘要】：
以大语言模型为代表的生成式人工智能正在迅速改变大学教学关于“人”及其教育的生产方式，在全球高等教育领域塑造出一种全新的人才培养模式。基于斯坦福大学的案例发现，大语言模型赋能教学的关键在于大数据集算力加速精准化教学，符号语言系统支持全时场服务，神经网络模型促进个性化指导以及智能情感技术实现人性化评价，产生了智能化涌现、无边界探索、生成式互动与情感化反馈的教学变革影响。但与此同时，人工智能“幻觉”、机器“认知偏见”与复杂系统“失控”的模型应用限度，又可能引发教学内容失真、意识形态危机与主体关系异化的多重风险，催生技术革命下新的教学危机。基于此，人工智能大模型时代的大学教学理应从“主体—目标—体系—机制”的四维关系构建系统路径，一体推进生成主义式教学创新探索、教学模型信用评级提升与多元协同教学生态建构，促进大学教学从人机协同走向人机融合，从而彰显大语言模型赋能大学教学的技术进步主义价值。
【英文摘要】：
Generative artificial intelligence, represented by large language models, is rapidly changing the production method of "human" and their education in university teaching, shaping a new model of talent training in the global higher education field. Based on the case of Stanford University, it is found that the key to empowering teaching with large language models lies in the computing power of large data sets to accelerate precise teaching, the symbolic language system to support full-time services, the neural network model to promote personalized guidance, and the intelligent emotional technology to achieve humanized evaluation, resulting in the teaching transformation impact characterized by intelligent emergence, boundless exploration, generative interaction and affective feedback. But at the same time, the limits of model application such as the "illusion" of artificial intelligence, the "cognitive bias" of machines, and the"out-of-control" of artificial intelligence may lead to multiple risks of distorted teaching content, ideological crisis and alienation of subject relations, giving rise to a new teaching crisis under the technological revolution. Based on this, universities in the era of artificial intelligence should build a systematic path from the four-dimensional relationship of"subject-goal-system-mechanism", promote generative teaching innovation and exploration, improve the credit rating of teaching models, and build a diversified collaborative teaching ecology, so as to change university teaching from human-computer collaboration to human-computer integration, thereby demonstrating the technological progressivism value of university teaching empowered by the large language models.