In this talk I will present a novel view on generative AI systems, how they challenge norms, and how we should relate to them. I argue that these systems are role-playing devices that excel at pretending they are knowledgeable. They do not have intentionality or conscience, but excel at reproducing style and structure. This is not necessarily a problem, provided we learn when and how and where to use these systems.

In the talk, I introduce the concept of epistemic capture as a heuristic to analyze whether these agentic systems are causing individual, social, or corporate harm, and how to prevent that capture from happening.