Designing a physical quantum agent

The concept of an embodied intelligent agent is a key in modern artificial intelligence and robotics. Physically, open system embedded environment that it interacts with through sensors actuators. It contains learning algorithm correlates the sensor actuator results by features about its environment. In this article we present simple optical uses light to probe learn components our scenario, quantum outperforms classical agent: probes world using single photon pulses, where counterpart weak coherent state average number equal one. We analyze thermodynamic behavior both agents, showing improving agent's estimate corresponds increase work done on pulse. Thus, model provides useful toy for studying interface between machine learning, optics, statistical thermodynamics.