A recent article in PNAS Nexus by Athanassios S. Fokas explores the potential of artificial intelligence (AI) to equal or surpass human cognitive abilities. Traditionally, AI is judged on tasks such as winning games or mimicking human conversations. However, Fokas highlights a challenge: testing AI on every conceivable human goal is not feasible.
He proposes the need for alternative methods. The common focus on “complex goals” fails to capture elements of human thought like emotion, subjective experience, and understanding. Additionally, AI lacks true creativity, unable to connect diverse topics using metaphor and imagination to generate novel outcomes not explicitly set as goals.
While AI models often mimic artificial neural networks, human thinking extends beyond neurons, involving the entire body and various brain cells like glia cells, which AI does not replicate. Fokas argues that computations only represent a small part of conscious thinking, with a significant amount of unconscious processing at play. Consequently, Fokas concludes that AI has a long way to go before surpassing human thought.
