Abstract

In the quantum science special exhibition at the National Museum of Natural Science (NMNS), Taiwan, visitors debate quantum mechanics with life-size AI digital twins of Einstein and Bohr, and step into an AI-animated photograph of the 1927 Solvay Conference. These installations are the most visible examples of the museum's work with AI, but this talk focuses on the broader institutional strategy behind them.

From 2023 to 2025, NMNS piloted AI across administration, curation, and visitor services, including meeting transcription, operational data analysis, voice reconstruction, and generative imagery. Two lessons emerged. A study of 268 visitors showed that the ability to change content does not by itself create a sense of participation; visitors need to see that their choices produce a meaningful response. Operational experience also showed that maintaining trust requires continuous staff attention through hallucination monitoring, content validation, and clear labeling of AI-generated interactions.

Building on these experiences and staff training that reached about 450 participants, NMNS launched a museum-wide AI empowerment program in 2026. Twenty bottom-up proposals from fourteen units are supported through consultation, a six-course curriculum, and shared computing and knowledge infrastructure.

The talk also considers unresolved questions, including rights over a scientist's digital identity and whether a digital Einstein should ever learn new physics. The guiding principle is that capacity, evidence, and trust must develop together if AI is to serve a museum's purposes.

Biography

· 2016–present, Assistant Curator, Science Education Division, National Museum of Natural Science, Taiwan
· 2014–2015, Researcher, Laboratoire d'Astrophysique de Marseille (LAM), France