Constitutions, AI, and the Meaning of Rules | Ep. 10 скачать в хорошем качестве

Constitutions, AI, and the Meaning of Rules | Ep. 10

In this conversation, I sit down with philosopher Dr. Gillman Payette to explore what “constitutions” really mean — for both humans and artificial intelligence. We start from first principles: why do humans create constitutions at all? Drawing on Hobbes, social contract theory, and the philosophy of law, we examine how constitutions clarify norms, define obligations, and provide structure to moral life. From there, we turn to Anthropic’s Constitutional AI, which tries to encode moral principles directly into the training process of large language models. Together we discuss its shortcomings and ambiguities — the vagueness of terms like “harmful,” “helpful,” or “good,” the dependence on human interpretation, and the lack of a clear semantic grounding compared with the precision of human legal documents. We compare this to human constitutions, which rely on interpreters, courts, and shared understandings, while AI constitutions depend on statistical language patterns and opaque moral averages learned from data. 🔗 Connect with Gillman Payette https://profiles.ucalgary.ca/gillman-... 🔗 Anthropic's Constitutional AI https://www.anthropic.com/research/co... 🔗 Mentioned article by Stephen Wolfram https://writings.stephenwolfram.com/2... 💡 Get Involved with the Podcast If you’d like to be a guest on Virtuous Machines, know someone who should be, or want to get involved with the podcast in other ways, reach out at: contact@virtuousmachines.com #ai #AIethics #ConstitutionalAI #PhilosophyOfAI #AIsafety #MoralPhilosophy #AIgovernance #Alignment #ArtificialIntelligence #EthicsInTech #VirtuousMachines