With a large dataset of generated policies, the next step is to import them back into the primary software application that displays the 360-degree video. This integration allows the dynamically generated rules to influence the visual output or behaviour of the system in real-time. My use of the term “policy” is a deliberate nod to its origins in the field of Reinforcement Learning (RL), a concept dating back to the 1990s. In RL, a policy is the strategy an agent employs to make decisions and take actions in its environment. It is the core component that dictates the agent’s behaviour as it learns through trial and error to maximise cumulative reward. By generating policies based on visual input, my system is, in a sense, creating its own world model—a simplified, learned representation of its environment and the relationships within it. This process echoes the fundamental principles of how an AI agent learns to react to and make sense of the real world, a topic I have delved into in more detail in some of my earlier writings.