How AI Is Preserving Holocaust Survivor Stories for Future Generations

In a museum in Illinois, a man named Pinchas Gutter answers questions from visitors. He tells them about his childhood in Poland, about his family, about the events of the Second World War. He responds to questions he has never been asked before, with answers that are specific, detailed, and emotionally alive. He is ninety-four years old. He is also, technically, not in the room — but a high-definition projection of him is, powered by an AI system trained on hours of recorded testimony.

Pinchas Gutter is one of several Holocaust survivors who have participated in the USC Shoah Foundation's Dimensions in Testimony project — one of the most ambitious and technically sophisticated legacy preservation efforts in history. The project records survivors over multiple days in a specially designed studio, asking them thousands of questions across every aspect of their experience and their lives. The recordings are then used to train an AI system that can respond to new questions in real time, drawing on the corpus of recorded answers to produce responses that are, to the greatest extent possible, in the survivor's own words and voice.

The goal is explicit: to allow children and grandchildren born after the last survivor has died to still have a conversation with one — to ask their own questions and receive answers that approximate what the survivor would have said.

The technology is not perfect. The AI cannot generate truly new information — it can only recombine and surface what was captured in the original recordings. But the original recordings are comprehensive enough that the recombinations feel, to most users, genuinely responsive. Visitors to the museum frequently report emotional reactions to the interaction that they did not expect — a sense of actual presence that photographs and even videos cannot produce.

The principle that drives the Dimensions in Testimony project is the same principle that drives Heirloom: the best time to capture a story is while the storyteller is alive, and the more comprehensively you capture it, the more it can do in the future. The technology will only improve. What it cannot do is reach backward in time to capture what was not recorded when the person was alive.