Sanyu Studio: AI System Aims to Pluralize Art-Historical Narratives
A recent study published on arXiv, titled 'Sanyu Studio: A Multi-Agent System for Art-Historical Narrative Construction,' addresses concerns regarding generative AI potentially leading to a uniform art history. The researchers created a dialogue-driven system that treats 321 oil paintings by Sanyu, a pioneer of Chinese-French modernism, as independent agents equipped with methods for fact-checking, interpretation, organization, and memory filtering. An intensive workshop lasting a week, involving eight art university students, demonstrated that the construction of narratives is shaped by user prompts, the arrangement of evidence, and cognitive biases, resulting in various interpretations of 'digital Sanyu.' The findings indicate that AI can amplify human creativity and expand art interpretation through interactive digital platforms. The complete study can be accessed on arXiv in the Computer Science > Artificial Intelligence category.
Key facts
- A new arXiv preprint proposes Sanyu Studio, a multi-agent dialogue system for art-historical narrative construction.
- The paper is filed under Computer Science > Artificial Intelligence.
- Sanyu Studio models 321 Sanyu oil paintings as agents.
- Each agent is equipped with fact, interpretation, organization, and memory-filtering mechanisms.
- The system was tested in a seven-day workshop with eight art-university participants.
- User prompts, evidence organization, and cognitive tendencies shaped distinct but internally consistent versions of 'digital Sanyu'.
- The authors argue that AI can amplify human agency and provide an interactive entry point for public audiences, especially when historical evidence is limited.
- The paper's full text, code, and data are linked on arXiv, with references to Semantic Scholar and arXivLabs.
Entities
Artists
Institutions
- arXiv
- Semantic Scholar