Abstract: Predators chasing prey, investors trading stocks, a patron playing blackjack: all can be thought of as agents that are required to make real-time decisions based on a combination of present observations and experience. As the tasks they seek to achieve become ever more complex, executing strategies that maintain consistency over ever-greater timescales and ever-greater awareness of their environment. Success thus requires such agents to maintain more complex models, tracking ever-growing amounts of data that incur additional computational or energetic resource costs.
Could such an agent gain an operational advantage by processing quantum information?
Here, I introduce a framework to describe quantum-enhanced agents - automated machines capable of processing data quantum mechanically. I show that such agents can leverage quantum information to track a less complex model of reality without necessary loss in adaptive function. I then outline the potential resource cost savings, enabling a fundamental reduction in energetic costs for quantum machines to respond to environmental stimuli in real time.
Keywords: IPW Intelligence Physical World FFF, IPW 2019, Gu