World Models and the Emergence of a “First-Person” Perspective in an AI

Could an artificial agent develop an operational first-person perspective through its own interaction history? This talk considers embodied systems that predict, receive feedback and integrate multiple modalities through global-workspace-like dynamics. Using the Conscious Turing Machine framework, it proposes that an agent-centred world model can acquire a temporally organized point of view. The discussion connects that structural proposal with enaction and neurophenomenology without establishing subjective experience in machines.