Extending Human Intelligence Through AI

Modern AI systems demonstrate remarkable capabilities in essay writing, code generation, and summarizing complex ideas, but they still show limitations in intuitive tasks such as object tracking, logical reasoning, and discerning truth from falsehood. This contradiction has sparked debate over whether AI should be viewed as an intelligence similar to humans, or as a sophisticated autocomplete function.

Recent research offers a new perspective: that AI does not simply replicate human intelligence, but instead leverages a system rooted in human cognitive structures. From a Phenomenology perspective, language already contains structures in which human understanding has accumulated, and AI operates by modeling and extending these structures.

Microsoft Research's paper, 'The Origins of Artificial Intelligence in Natural Intelligence', argues that rather than defining AI as either a human mind or a mere statistical technique, it should be understood as a tool that extends structures originating from human cognition.

This perspective also changes the approach to AI Safety. Rather than framing AI through the narrative of a 'rogue AI,' the focus should shift to building trustworthy AI through system-level engineering and governance.