Zenith Grant Awardee

Thomas Elliott

Nanyang Technological University

Co-Investigators

Jayne Thompson, Centre for Quantum Technologies; Felix Binder, Nanyang Technological University

Project Title

The role of quantum effects in simplifying adaptive agents

Project Summary

Whether it's a predator tracking their next meal, or a seed waiting for the right conditions to blossom, the key to survival in nature lies in making appropriate decisions based on present stimuli and past experiences. All living things continuously monitor their environment, devoting resources to store and track relevant information so that they can anticipate how each of their actions might affect the future. Interactions between such agents drives an evolutionary arms race, where they must devote ever-increasing resources to survive in increasingly complex environments. Could the drive for survival motivate adaptive agents to harness quantum effects? In this project, we approach this question by interfacing elements of complexity and quantum science. The former allows us to identify the resource costs for agents to adapt to various complex environments, and the latter to reduce this cost below classical limits when such agents are capable of processing quantum information. Such discoveries can have profound consequences: they uncover the means for designing quantum machines that can adapt to and thrive in their environment, while simultaneously illustrating that evolutionary pressure may well favour life capable of quantum reasoning.

Technical Abstract

An agent is an entity that has the capacity to influence its environment. They can appear in many guises: a predator hunting in the wild; a car navigating through traffic; a seed looking to germinate. From an information-theoretic perspective, these situations are all described within a unified framework, input-output processes, where input data is transduced into an output sequence. An agent receives stimuli (inputs) in the form of observations, and based on these will execute a strategy of actions (outputs) in response. Effective strategies to execute in response to complex environments must themselves too be typically complex. This requires agents with the ability to adroitly anticipate and adapt to their surroundings to track large amounts of information. This drives an evolutionary pressure for agents to adopt structures that can efficiently compress the information their strategy must track. We ask whether this drive might motivate adaptive agents to harness quantum effects. This is motivated by previous work on memory-efficient quantum models of stochastic data sequences. Using a blend of tools from quantum and complexity science, we will determine the most efficient memory structures for classical agents, and corresponding structures for quantum agents that offer compression beyond classical limits.