Dualism

Blog · Read

• Uses the “conjurer’s trick” of magic-trick metaphors to argue much of the AI bubble rests on conflating different processes that merely produce similar-looking outcomes.
• Distinguishes relationship-based understanding from statistical prediction: both can guess correctly what someone will say, but only understanding provides a path forward when the guess fails — statistical extrapolation just “founders.”
• Revives the degraded Turing Test critique: the original thought experiment has been flattened into “can a chatbot fool a human,” which is a different question from whether something is actually a person.
• Applies the same dualism to labor — “centaurs” (workers voluntarily adopting AI tools) versus “reverse centaurs” (workers conscripted as peripherals for machines) — and to job losses: genuine AI-driven replacement versus a boss merely swindled into believing a defective chatbot could replace someone.