Peking University debuts Luo Shenfu 2.0 aesthetic AI creator

Luo Shenfu 2.0 is a good example of how cultural institutions are starting to treat AI not just as a content engine, but as a carrier for civilizational values. By encoding a large canon of Chinese aesthetic theory into a structured “gene bank” that models concepts like “grandeur”, “melancholy” or “ethereal” across text and imagery, Peking University is building a specialized foundation layer that could sit underneath many creative agents. This is narrow AI, but it coexists with general‑purpose models and shows how cultural priors can be made explicit rather than left to opaque weights learned mostly from Western internet data.

For the AGI race, this kind of work matters because it broadens the notion of what “intelligence” should align to. As more countries build cultural AI stacks—codifying local aesthetics, values, and metaphors—the frontier systems that succeed globally will need to interoperate with, or subsume, these domain ontologies. That’s likely to push AGI research toward models that can ingest structured cultural knowledge bases alongside unstructured data, and that can generate outputs that feel native to different aesthetic and philosophical traditions, not just English‑language internet culture.