AI microdramas cause ‘aesthetic fatigue’, professor tells Macau forum

A Beijing professor has cautioned at a Macau forum that formulaic AI-generated microdramas could stifle human creativity and suppress artistic innovation. Zhang Lanshan, a professor at the Beijing University of Posts and Telecommunications' School of Digital Media and Design Arts, warned that while audiences enjoy standardized content, it poses a threat to diversity and artistic endeavor.

Experts from Hong Kong institutions echoed her concerns about cultural homogenization and disinformation that large language models may bring.

Microdramas are short, social media-friendly video clips, usually lasting between 60 to 120 seconds. Zhang explained that AI models like Jimeng and Xiaoyunque are trained on data from these viral productions. The opening six seconds often contain a hook to grab the audience's attention, while the next 30 seconds deliver a thrilling plot twist at the one-minute mark.

While this formulaic approach generates substantial traffic, Zhang argues it can lead to homogenization and diminish the diversity and innovation of creative work. She further stated that these productions lack humanistic values and endurance, being inferior compared to human-created content.

Zhang advised people to maintain their independent judgment and active decision-making process to prevent manipulation, emphasizing that AI should serve as a tool to augment creativity rather than being the sole driver of production. Wu Qiong, director of Tsinghua University's Academy of Arts and Design, also addressed the issue, warning about the potential loss of cultural differences and misconceptions that AI's focus on common denominators and averages could create.

According to Wu, Chinese accounts for only 1.1 percent of online content globally, as opposed to 49.2 percent for English, debunking the notion that AI training data predominantly consists of Chinese.

Zhao Hai, vice-president of the Chinese Institute of Hong Kong, warned users about AI's risks, despite its convenience in daily life, such as eliminating language barriers. Zhao pointed out that AI's answers might occasionally seem off due to discrepancies between its training data and human knowledge, which could reflect cultural or civilizational differences.

However, Zhao also cautioned that AI's output might be biased or manipulated through extensive training, potentially leading to disinformation or misinformation, particularly in Western societies. He emphasized the importance of preserving cultural diversity and adopting human-centered AI ethics to counteract these concerns, acknowledging that technological advancements may not necessarily promote human empathy.