Uncanny Testimony (longreads.com)

Museums and nonprofits are using AI to create interactive, “immortal” testimonies of Holocaust survivors: projects like USC Shoah Foundation’s Dimensions in Testimony project a life‑size hologram that answers visitor questions by selecting pre‑recorded clips via an AI algorithm (researchers found visitors tend to ask the same ~100 questions). Similar initiatives include the Museum of Jewish Heritage’s Survivor Stories, Testimony 360 VR tours, Liberation75’s AI de‑aging of survivors’ videos, and chatbots trained on historic writings (e.g., SchoolAI and yeschat.ai). Interfaces range from in‑theater holograms to web chat and VR; the underlying tech is less generative voice‑synthesis and more retrieval+selection from curated archives, sometimes augmented with NLP to match questions to recorded responses. The shift matters because it preserves dialogue after survivors die, expanding access and pedagogical reach—but it also raises technical and ethical tensions. Retrieval‑based systems can replicate answers verbatim and flatten nuance, while chatbot or de‑aged outputs risk inventing or sanitizing material and overstating authenticity. For archivists and AI practitioners this highlights tradeoffs between fidelity, scalability, and user expectation: robust indexing, provenance metadata, transparent limits of historicity, and careful UI design are critical to prevent harm. The story underscores a broader point for the AI/ML community—technical choices in search, retrieval, and synthesis carry profound moral weight when they stand in for real human memory.