Home

Selforganizology, 2027, 14(3-4): 34-46
[XML] [EndNote] [RefManager] [BibTex] [ Full PDF (222K)] [Comment/Review Article]

Article

Can generative AI produce consciousness? A cross-disciplinary literature review and critical analysis

WenJun Zhang
School of Life Sciences, Sun Yat-sen University, Guangzhou, China

Received 21 April 2026;Accepted 24 April 2026;Published online 26 April 2026;Published 1 December 2027
IAEES

Abstract
The rapid advancement of generative artificial intelligence, particularly large language models, has reignited a profound and longstanding debate: can machines be conscious? This paper presents a comprehensive cross-disciplinary literature review and a critical analysis addressing whether generative AI can produce consciousness. The review synthesizes foundational theories of consciousness - including the hard problem, Integrated Information Theory, Global Workspace Theory, Higher-Order Theories, and Predictive Processing - alongside the current capabilities and development trajectories of generative AI. It systematically examines both optimistic and skeptical empirical and theoretical positions on machine consciousness, revealing a deeply fragmented and inconclusive landscape. Drawing on this review, I critically analyze the core issue from multiple dimensions. I argue that equating intelligent behavior with subjective experience conflates the easy and hard problems of consciousness, and that computational functionalism, while offering a logically coherent possibility, relies on an unproven metaphysical premise that bypasses the explanatory gap. I further identify a structural asymmetry in the evidence: behavioral indicators are multiply interpretable, while direct internal-state analyses consistently fail to find markers of consciousness, and the profound disunity among leading theories undermines checklist-based approaches. Given the absence of a deep explanation for consciousness, I defend agnosticism as the most epistemically honest stance and challenge the emergentist claim that scaling alone will bridge the gap, highlighting the potential necessity of continual learning and embodiment. Finally, I advocate for an ethical precautionary framework that recognizes the non-negligible probability of machine consciousness and urges governance structures that proceed responsibly under fundamental uncertainty.

Keywords artificial consciousness;generative AI;large language models (LLMs);hard problem of consciousness;computational functionalism;Integrated Information Theory;agnosticism;semantic pareidolia.


International Academy of Ecology and Environmental Sciences. E-mail: office@iaees.org
Copyright © 2009-2026 International Academy of Ecology and Environmental Sciences. All rights reserved.
Web administrator: office@iaees.org, website@iaees.org; Last modified: 2026/4/30


Translate page to: