Article Details
Comparative Analysis of Theories towards Artificial Consciousness Author Name: Ms Anastasia Drikakis Co-author Name: Stavroula Tsinorema Abstract: This paper aims to review some of the dominant philosophical, mathematical and neuroscientific theories surrounding consciousness and their essential properties. After establishing the conditions in which consciousness can be attributed, it will do so by briefly providing a comparative analysis of Integrated Information Theory (IIT), Global Neuronal Workspace (GNW) and Higher-Order theories of Consciousness amongst others1, and present our own thesis with robust argumentation. Finally, it will consider implications of these theoretical frameworks for understanding and replicating consciousness in artificial systems. Specifically, drawing upon the transformative capabilities of LLMs such as GPT (Generative Pre-trained Transformer) models, it will examine the potential for these systems to exhibit characteristics indicative of consciousness1. By examining experiments assessing large language models’ understanding of self-referential concepts, empathy, and ethical reasoning, the paper concludes despite their capacity to engage in introspective tasks and exhibit signs of autonomy, consciousness requires properties above and beyond those fostered by current neural network models2. This is as all theories discussed require subjective emergence, an attribute not possessed by current AI. Keywords: Consciousness, Artificial Intelligence, Large Language Models, Philosophy of Mind. Topics: ENGINEERING Subtopic: Artificial Intelligence

Article Details

Comparative Analysis of Theories towards Artificial Consciousness

Author Name:

Ms Anastasia Drikakis

Co-author Name:

Stavroula Tsinorema

Abstract:

This paper aims to review some of the dominant philosophical, mathematical and neuroscientific theories surrounding consciousness and their essential properties. After establishing the conditions in which consciousness can be attributed, it will do so by briefly providing a comparative analysis of Integrated Information Theory (IIT), Global Neuronal Workspace (GNW) and Higher-Order theories of Consciousness amongst others1, and present our own thesis with robust argumentation. Finally, it will consider implications of these theoretical frameworks for understanding and replicating consciousness in artificial systems. Specifically, drawing upon the transformative capabilities of LLMs such as GPT (Generative Pre-trained Transformer) models, it will examine the potential for these systems to exhibit characteristics indicative of consciousness1. By examining experiments assessing large language models’ understanding of self-referential concepts, empathy, and ethical reasoning, the paper concludes despite their capacity to engage in introspective tasks and exhibit signs of autonomy, consciousness requires properties above and beyond those fostered by current neural network models2. This is as all theories discussed require subjective emergence, an attribute not possessed by current AI.

Keywords:

Consciousness, Artificial Intelligence, Large Language Models, Philosophy of Mind.

Topics:

ENGINEERING

Subtopic:

Artificial Intelligence