Active Inference: from Neuroscience to Real World Systems

Abstract

The Free Energy Principle, has been proposed as an all-purpose model of the brain and human behavior that crucially closed the loo p with action informing inference. As a formal and technical first principles mathematical account of how brains work, it has garn ered increasing attention from computer science to philosophy. The theory is based on the mathematical formulation of surprise minim ization, to do so a brain can minimize its Free Energy (a computable bound on surprise), and drive, not only perception and cognitio n but crucially also actions. As a framework, the Free Energy Principle and its corollary Active Inference thus represents a funda mental departure from current systems in Artificial Intelligence, as it calls for the implementation of a top-down system, rather th an a bottom-up system (driven by masses of training data) that are currently the state-of-the-art frameworks in AI research. This pr oposal aims to deliver a software and simulation platform that can be used to apply the Free Energy Principle and Active Inference a s an AI solution to simulated real-world problems.

Document Details

Document Type
DoD Grant Award
Publication Date
Aug 20, 2021
Source ID
N629092112043

Entities

People

  • Biswa Sengupta

Organizations

  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Neural Network Machine Learning.
  • Snow Cover Descriptors for Reptiles and Their Illustrations.
  • Systems Analysis and Design

Technology Areas

  • AI & ML
  • AI & ML - DoD AI Strategy
  • AI & ML - Machine Learning Algorithms