Making the Thermodynamic Cost of Active Inference Explicit.

Area of Science:

• Theoretical neuroscience
• Statistical physics
• Quantum information theory

Background:

• Active Inference Agents (AIAs) are models of intelligent systems.
• The term "energy" has dual meanings in AIAs: physical energy utilization and statistical Variational Free Energy (VFE).
• Understanding the interplay between these energy concepts is crucial for explaining organismal behavior and survival.

Purpose of the Study:

• To develop a theoretical account of Thermodynamic Free Energy (TFE) in Active Inference Agents (AIAs).
• To elucidate the relationship between TFE and Variational Free Energy (VFE).
• To explore the macroscopic consequences of tradeoffs between TFE and VFE for organismal strategies.

Main Methods:

• Formulation of Thermodynamic Free Energy (TFE) within a quantum information-theoretic framework.
• Analysis of the necessary tradeoffs between TFE and VFE.
• Connecting these theoretical tradeoffs to macroscopic biological strategies.

Main Results:

• A unified account of physical energy (TFE) and statistical energy (VFE) in Active Inference Agents (AIAs).
• Identification of inherent tradeoffs between energy utilization and information processing in biological systems.
• Demonstration of how these tradeoffs underpin diverse metabolic strategies across organisms.

Conclusions:

• The explicit formulation of TFE and its relationship with VFE provides a theoretical basis for understanding biological energy management.
• These findings offer insights into the evolution of metabolic strategies, from plants to predators.
• This work bridges concepts from physics, information theory, and neuroscience to explain fundamental aspects of life.