Daoxin Li of Peking University presented his recent project combining the history of cinema and AI.

Last Wednesday, Professor Daoxin Li presented “Rethinking Film and Film History in the Age of Digital Intelligence,” a discussion on his research in artificial intelligence and Chinese film history, which was particularly fascinating for a Computer Science and Film Studies major like yours truly. He began by recounting his past experiences with Columbia University, having delivered several talks here before. The lecture was given entirely in Mandarin. Thankfully, for those (including this writer) who are sadly not blessed with an understanding of the language, a post-doc affiliate provided English translation, seemingly with AI assistance. He presented the trailer for Bi Gan’s Resurrection (2025), which he described as a “meta” film exploring the human senses. IMDb describes the plot as follows: “In a society where people stop dreaming to extend their lifespan, some dangerous individuals still dream, warping the fabric of time. We experience five dreams, for each of the senses, each chronologically representing a period of cinema.” Professor Li named Bi Gan as part of a new generation of Chinese directors, which in turn are part of a new generation of cinema itself.

With the capabilities of artificial intelligence rapidly progressing, including image and video generation, Li sees this technology as changing the scope of cinema. To illustrate this concept, he showed a photo from a clip of Brad Pitt fighting Tom Cruise, a scene created entirely using an AI program called Seedance 2.0. He then posed the question, “what is cinema?” This further gave way to “what is cinema history?” Li views his new platform, Chinese Cinema Knowledge System (CCKS), as transitioning from rewriting cinema history to conceptually reconstructing cinema. He delineated the history of film into three parts, beginning with the celluloid era of the past century. Harkening back to the themes of Resurrection, he explained how film creates a bridge between reality and dreams. The images of this era are reminiscent of a dreamlike state, epitomized by darkrooms, projectors, and film reels, which evoke the once-solid materiality and mystery of classic cinema as part of everyday life and collective memory. The next era, the present, is the digital age, one of ontological uncertainty. The act of viewing film has drastically changed with the ability to experience it outside of the cinema through streaming, removed from its prior dreamlike environment. This brought him to the final era, where the cinema screen becomes the world itself, everything will be computable, and film may lose its typical subject-object relationship, dissolving instead into a flow of information. Would cinema disappear in such an era?

Under this framework, Li claims the history of film can now be thought of as decentralized, uncertain, and nonlinear. The CCKS platform is his answer to this erosion, based on a multimodal language model using vast historical datasets, and introduced by Professor Li through an AI-generated video. Li works as a Vice Dean, Doctoral Supervisor, and Professor in the Department of Drama and Film Studies in the School of Arts at Peking University, where an ongoing institutional research goal is integrating digital technology and the humanities. This led him to develop CCKS as a cross-media transformation of academic resources. CCKS is a high quality, data-driven, and autonomous knowledge system on Chinese cinema history. Li emphasized the necessity of focusing on Chinese cinema and enhancing its availability of resources, citing the influence of Western theories and ideas even at the academic level in discussions of the subject. In making CCKS, he strives to create a film platform with distinct Chinese characteristics through cross-disciplinary incorporation of AI and filmmaking. Li explained that he himself was among the first, presumably in academic history, to study film and film history at a university level.

Many current databases include word-of-mouth or inaccurate information, which CCKS aims to combat through primary sources. Using tools like Chinese Character Tracking and Observation Platform (CTOP), written materials such as newspapers or handwritten letters can be digitized; CCKS utilizes tens of thousands of documents. Additionally, Li cooperated with 1905 Film Network under China Movie Channel (CCTV-6) and the hometown of Zheng Zhengqiu (considered the founding father of Chinese cinema) in creating CCKS. His other connections include an international summer school program hosted for the development of CCKS and a film about film history, specifically filmmaker Fei Mu, for which Li provided consultation. Li hopes CCKS’s new approach can be used to create a solid digital foundation for cinema.

Perhaps the most impressive feature of CCKS is the Cinemetrics tool. Li demonstrated Cinemetrics through a prepared video. The tool takes in footage and breaks down its shot structure, transcribes dialogue, and provides color analysis. Cinemetrics separates a clip into its individual shots and returns data on the type of shot (such as close, medium, or wide) and its duration. It is able to map the number and length of shot types present in a given video and supply word cloud visualization for its dialogue. This allows entire films to be dissected instantaneously for quantitative interpretation, also allowing the user to conceptualize trends in filmmaking styles throughout history.

Li showed a portion of Laborer’s Love (1922), the first known complete Chinese film that is still available. Multiple versions of the film exist today. For further research, Li wants to examine how landmarks in cinema history could be reconstructed using CCKS, with Laborer’s Love as an example. CCKS has yet to gain recognition on a public or industry level, and is presently known primarily among academics.

One student’s question regarded the possibility of CCKS influencing an artistic homogenization in filmmaking, since it employs AI and could be used to evaluate popular films or social media clips in order to generate similar content. Li answered that his goal is more oriented towards analyzing and comparing quantitative data metrics over time. This was the only concern addressed regarding the AI-powered tool; Li sees artificial intelligence as part of the future of cinema and cinema history, describing digital trends as “unstoppable.” Perhaps ironic to call upon considering Li’s emphasis on a non-Western focus, but the American perspective on AI is less enthusiastic, especially regarding its potential for negative effects on creative pursuits. Many risks are currently recognized by our own National Telecommunications and Information Administration (NITA), not to mention the environmental concerns and the technology’s tendency to “hallucinate” (refer to Bwog’s coverage of OpenAI). However, this was likely not an intentional omission, accounting for differences in international perception of artificial intelligence.

The evening concluded with Li’s attention to the transcendental nature of his research, followed by a reception in the next room featuring an extremely tempting charcuterie spread. Li underscored the significance of his work’s ability to possibly surpass a number of boundaries, including those between the past and present, China and the West, knowledge and action, and theory and practice. He aspires to witness the impact of CCKS in the reconstruction of film studies at the level of subject and object, with an equal importance placed on Dao and Qi (Daoist concepts relating to the universal path and guiding energy).

The scope of Professor Li’s research is undoubtedly impressive. Despite personal reservations about the use of artificial intelligence, which may be best exemplified by one instance during the lecture where the postdoc began to struggle in providing translation for Li and a nearby guest suggested she do so directly rather than “with the AI,” this writer does find his project innovative and remarkable. It would certainly be intriguing to see how CCKS evolves with Li’s further development.

Image from CCKS website via the Author