Author(s) :

Jiarui Liu.

Volume/Issue :

Abstract :

Despite claims that escalating computational complexity and sophisticated learning architectures could potentially enable AI systems to attain self-awareness, numerous scholars and technologists maintain that, regardless of the advancement in their linguistic proficiency and contextual comprehension, Large Language Models (LLMs) inherently lack the subjective experiences essential for authentic self-consciousness. The article explores the philosophical underpinnings that differentiate basic awareness, an entity’s ability to process and react to environmental cues, from higher-order self reflection, which involves an agent’s introspective recognition and contemplation of its own mental states. In the paper, it has been analyzed the central theoretical constructs proposed by Thomas Nagel, José Luis Bermúdez, and David Rosenthal. Moreover, the paper employs the red dot (mirror) test and consider digital adaptations to assess LLM’s detection of hidden changes in their outputs. LLMs exhibit reliable recall of previous responses within a single session; however, their episodic retrieval lacks the necessary continuity to sustain personal identity across interactions. Ultimately, despite their advanced functionalities, LLMs are devoid of authentic subjective experience and individuality. The separation of simulated self-awareness from genuine consciousness is essential for the responsible progression of AI. Philosophical delineations are crucial for distinguishing simulations from real experiences, a factor vital for ethical guidelines and policy frameworks in AI’s societal integration. Philosophical delineations are crucial for distinguishing simulations from real experiences, a factor vital for ethical guidelines and policy frameworks in AI’s societal integration. The objective of this paper is to examine whether Large Language Models could develop self-consciousness by analyzing philosophical definitions, reviewing current cognitive science perspectives, and evaluating experimental approaches designed to test machine awareness.