Machine Consciousness and the Global Workspace Theory

Abstract

The past century has seen a resurgence of interest in the study of consciousness among scholars of various fields, from philosophy to psychology and neuroscience. Since the birth of Artificial Intelligence in the 1950s, the study of consciousness in machines has received an increasing amount of attention in computer science that gave rise to the new field of machine consciousness (MC). Meanwhile, interdisciplinary research in philosophy, neuroscience, and cognitive science has advanced neurocognitive theories for consciousness. Among many models proposed for consciousness, the Global Workspace Theory (GWT) is a promising theory of consciousness that has received a staggering amount of philosophical and empirical support in the past decades. This dissertation discusses the GWT and its potentials for MC from a mechanistic point of view. To do so, Chapter 1 gives an overview of the philosophical study of consciousness and the history of MC. Then, in Chapter 2, mechanistic explanations and tri-level models are introduced, which provide a robust framework to construct and assay various theories of consciousness. In Chapter 3, neural correlates (and thereby, neurocognitive theories) of consciousness are introduced. This chapter presents the GWT in details and, along with its strengths, discusses the philosophical issues it raises. Chapter 4 addresses two computational implementations of the GWT (viz., IDA and LIDA) which satisfy specific goals of MC. Finally, in Chapter 5, one of the philosophical problems of MC, namely, the Frame Problem (FP), is introduced. It is argued that the architectures based on the GWT are immune to the FP. The chapter concludes that the GWT is capable of "solving" the FP, and discusses its implications for MC and the computational theory of mind. Chapter 6 wraps up the dissertation by reviewing the content.

Does Machine Understanding Require Consciousness?

“…There is also an active line of research into designing systems capable of human-like faculties of perception, cognition and consciousness that is directly inspired by current neuroscientific theories of brain function ( Marblestone et al, 2016 ). Prominent among these are models based on the Global Neuronal Workspace (GNW) theory cited above ( Haqiqatkhah, 2019 ; Mallakin, 2019 ; Safron, 2020 ; VanRullen and Kanai, 2021 ). According to this theory, the brain contains many processes that are highly differentiated, localised, widely distributed and yet unconscious.…”