Researchers check digitalized mural images inside a Mogao cave in Dunhuang, Gansu Province. Photo: IC PHOTO

On Jan. 3–4, 2025, scholars from such fields as archaeology, cultural heritage studies, computer science, mathematics, and big data gathered in Zhuhai, south China’s Guangdong Province, to attend a forum themed “The Future of the Past: Charting the Frontier of AI in Archaeology and Cultural Studies,” where they explored new opportunities for integrating AI with these disciplines. 

Methodological innovation 

The past decade has witnessed rapid development of AI. Deep learning, in particular, has increasingly gained favor among archaeologists for its “superpowers” in pattern recognition, predictive analysis, and image processing. Chen Zhi, president of Beijing Normal University-Hong Kong Baptist University United International College (UIC), noted that as AI continues to integrate more deeply with archaeology and cultural studies, it will drive the emergence of new research outcomes and application scenarios. This, in turn, will inject sustained vitality into the preservation of humanity’s precious cultural heritage and the ongoing transmission of historical civilizations. 

Underpinned by the three pillars of algorithms, data, and computing power, AI has significantly enhanced research efficiency by enabling the creation of high-quality, diverse, and richly detailed databases. Ruan Yongbin, an academician from the Chinese Academy of Sciences and professor from the Institute for Advanced Study in Mathematics at Zhejiang University, examined AI’s distinct advantages in analysing massive and diverse archaeological data. 

Taking the periodization model for Neolithic Baodun Culture pottery fragments as an example, Ruan demonstrated the outstanding performance of AI in classification, clustering, and multimodal analysis, offering novel perspectives for scientific pottery periodization and cultural dissemination. He also predicted that AI will revolutionize archaeological methodologies, opening up new pathways for intelligent and precise studies. 

Zhichun Jing, a professor from the University of British Columbia in Canada, observed that deep learning has provided new methodological support for classification, a core task in archaeology. Classification is essential in archaeological research, whether for artifact typology, settlement pattern studies, or analysis of ancient social structures. Due to the large number of pottery fragments unearthed, traditional manual methods of reassembling them are inefficient. 

According to Fang Hui, director of the Institute for Cultural Heritage at Shandong University (SDU), his research team used AI-based reassembly techniques to recover many complete pottery objects from the Daxinzhuang site in Jinan, Shandong, significantly improving the efficiency of classification.

Empowering heritage protection

The application of science and technology has also demonstrated immense potential in the preservation and research of cultural heritage. Liu Cheng, a professor from the School of Cultural Heritage at Northwest University, shed light on specific application of digital technology in cultural relic restoration and preservation. For example, by combining 3D modeling with artifact analysis techniques, digital restoration and the recreation of ancient production processes have been achieved. Additionally, technologies such as digital photography, hyperspectral imaging, and infrared imaging have enabled the identification of damage zones and the revelation of hidden information in artifacts like cave temples, murals, rock art, and bronzeware.

Despite the progress, cultural heritage digitalization still faces numerous challenges, such as the absence of a unified cultural heritage information model, authenticity and process issues in modeling, insufficient expression of intrinsic value, and a lack of scientific rigor in digital monitoring. In response, Hu Di, an associate professor from the School of Geography at Nanjing Normal University, shared solutions from geographical perspectives, outlining four aspects of how digitalization can empower cultural heritage protection: comprehensive information acquisition and aggregation of cultural heritage, authenticity modeling based on specific contexts, digital twin expression of cultural heritage with integrated information, and geographic intelligent monitoring and simulation of cultural heritage.

In addition, heritage data is unique and difficult to obtain in large amounts, posing challenges to the prediction and generation of AI models. Zhuang Yiren, dean of the School of Culture and Creativity at the UIC, proposed a new strategy to collect data for 3D cultural heritage through crowdsourcing. His report detailed the workflow of capturing and completing 3D heritage models with the participation of volunteers to automatically generate additional data for AI data reinforcement. 

The forum was co-hosted by the Institute for Advanced Study at the UIC and the School of Archaeology at the SDU.

Edited by CHEN MIRONG