In a recent article published in the journal PNAS Nexus, Athanassios S. Fokas delves into a pressing question: Can artificial intelligence (AI) attain and surpass the complexity of human thought?
Traditionally, researchers have assessed AI capabilities based on its success in achieving intricate objectives, whether it be mastering games like Go or engaging in conversations indistinguishable from human interaction. Fokas, however, highlights a significant methodological constraint in this approach. To assert that an AI system is on par with human thought, it would need to undergo testing across the entirety of conceivable human goals, a formidable challenge.
Fokas proposes the necessity for alternative methodologies. The conventional focus on "complex goals" fails to encompass essential aspects of human thought, such as emotion, subjective experiences, and understanding.
Moreover, Fokas contends that AI lacks true creativity. Unlike humans, AI struggles to establish connections between vastly different subjects, employing methods like metaphor and imagination to produce innovative outcomes beyond explicit goals.
While AI models are often conceived as artificial neural networks, human thinking transcends mere neuronal activity. Human cognition involves the entire body and various types of brain cells, including non-neuronal glia cells.
Fokas emphasizes that computational processes represent only a fraction of conscious thinking, with conscious thought itself constituting just one facet of human cognition. A substantial amount of unconscious processing occurs behind the scenes. Consequently, Fokas concludes that AI is far from surpassing humans in cognitive abilities.
