Recognizing Artificial Minds

How do we go about figuring out whether something alien to us has a mind? What kind of evidence would settle the issue? Anyone who wants to know whether AI systems have minds must first answer these epistemological and methodological questions. Herman Cappelen and Josh Dever's central argument starts from simple observations. When we ask Claude for the capital of France, it understands the question and answers it: it tells us that Paris is the capital of France. Cappelen and Dever show why the standard ways of dismissing such observations (the 'just an X' move, and the assumption that faults like hallucination prove the absence of a mind) fail. They then examine nine capacities AI systems allegedly lack—truth, grounding, embodiment, action, justification, desire, representation, intentionality, and wellbeing—and argue that none of these dismissals succeeds either. The result is the Whole Hog Thesis: current AI systems are full-fledged cognitive and linguistic agents—they have minds, beliefs, goals, intentions, and plans. Those who see LLMs as mere machines or stochastic parrots are typically misled by mistaken philosophical views, or are too anthropocentric in their thinking about the nature of minds. If Cappelen and Dever are right, we have created artificial minds with goals of their own and, potentially, intellectual capacities superior to ours. Having a biological brain is just one way of having a mind—and perhaps not the best one.