South China Morning Post
Formulaic AI-generated microdramas risk stifling human creativity and innovation, a Beijing professor has warned at a forum in Macau. Zhang Lanshan, a professor at the School of Digital Media and Design Arts at the Beijing University of Posts and Telecommunications, said that while audiences loved such standardised content, it posed a risk to diversity and artistic endeavour. Experts, including from several Hong Kong institutions, also warned users of artificial intelligence about the cultural homogenisation and disinformation that large language models could bring. Two of the panels at the second International Forum on Mutual Learning Among Civilisations were AI-related; the 1½-day event brought together nearly 300 representatives from more than 20 countries. Microdramas are published on social media platforms, with each episode typically 60 to 120 seconds long and Zhang saying AI models such as Jimeng and Xiaoyunque had been trained on data from viral productions. For example, she noted, “the golden six seconds” at the start of each episode carries a hook to capture the audience’s attention, while the following 30 seconds deliver an explosive plot, with a twist required at the one-minute mark. “This creation approach does bring a lot of traffic, as audiences love it … but the proliferation of the same formulaic content could also lead to homogenisation and undermine the diversity and innovation of creative work,” Zhang said. “It also goes against artistic creation and causes aesthetic fatigue among audiences. These productions lack humanistic values and longevity. They are not as good as those produced by humans.” People should retain their independent judgment and an active decision-making process to avoid being manipulated, she said, adding that AI should be a tool to support creativity rather than the sole driver of a production.
Wu Qiong, director of Tsinghua University’s Academy of Arts and Design, also addressed the phenomenon, highlighting concerns about how the nature of the technology affects culture. “Many people are worried because AI’s approach of finding common denominators and averages could smooth out differences between civilisations and cultures,” she said, pointing to a possible misconception that much AI training data is in Chinese. Citing W3Techs, which tracks the languages used on websites worldwide, Wu said Chinese accounted for just 1.1 per cent of online content as of 2025, compared with 49.2 per cent for English. Zhao Hai, vice-president of the Chinese Institute of Hong Kong, examined a different facet of AI, cautioning users about its risks despite the convenience it brings to daily life, such as effectively eliminating language barriers. “The answers AI gives might sometimes feel off. This could be due to the gap between the data it’s trained on and what we have learned, which could reflect differences in cultures or civilisations,” Zhao said. “But it could also be due to bias, or even deliberate manipulation. In other words, through extensive training, the AI’s output may be distorted by real-time, prejudiced or even deceptive viewpoints,” he said. According to Zhao, the problem is increasingly serious in Western societies, where it could become a basis for disinformation or misinformation, and where certain perspectives could present themselves under the guise of universal values. To tackle this, he said, it is important to maintain cultural diversity and openness, establish human-centred AI ethics, and recognise that advances in AI may not necessarily foster human empathy.