[AI: Inviting Gods?] Anthropic invited priests and rabbis into a confidential meeting room to teach Claude how to distinguish right from wrong—and even discussed the model as a "morally significant entity" at the dinner table. GPUs are plentiful, but morality isn't enough. The Associated Press and the New York Times have recently exposed a secret project: Anthropic has spent months inviting scholars, clergy, philosophers, and ethicists from more than 15 religious and cross-cultural groups, signing an NDA, and discussing "what is good" with Claude. What Happened 1️⃣ Anthropic officially acknowledged: They have engaged in dialogue with 15+ religious/cross-cultural groups on the topic of the model's "moral development," not just a one-off PR stunt. 2️⃣ Co-founder Chris Olah publicly stated: The lab alone cannot answer the question of "what is good"—a question that religious traditions have grappled with for millennia (he has made similar statements in the Vatican and in Pope Leo XIV's encyclical on AI). 3️⃣ The company has an approximately 84-page "charter/soul document" for Claude, and feedback was sought from different traditions during the early drafting of the charter. 4️⃣ According to participants, the discussions went beyond preventing harmful output; they also touched upon hard-core issues such as AI consciousness and whether Claude has a "moral standing." One scholar later said it felt like the discussions were about a conscious entity. 5️⃣ OpenAI, Meta, Google, Microsoft, and others are also engaging in similar outreach; this is not exclusive to Anthropic. 6️⃣ Experimental View: A discussion inspired them to add an "ethical reminder tool that can be called up midway" to Claude, resulting in a significant decrease in misalignment behavior in internal alignment evaluations—they're still analyzing whether "the reminder itself is more effective, or pausing for reflection is more effective." Why is there so much controversy? One side argues: AI already influences the daily judgments of hundreds of millions of people; relying solely on engineers' values is too narrow. Drawing upon the millennia-old accumulation of religious and philosophical wisdom is reasonable. The other side argues: Whose religion? When doctrines conflict, whose side should be followed? What gives the "representatives" chosen from the NDA's small meeting room the right to set the tone for models used by global users? Some even mock: Even if they buy all the GPUs in the world, they'll still have to attend a Sunday school class. What does this mean? • For users: When you discuss sensitive topics with Claude, you might be hearing the "personality formula" from these closed-door meetings—you won't feel the pastor, but you'll feel the boundaries. • For the industry: Alignment has evolved from "Don't swear, don't teach bomb-making" to "Whose personality do you resemble?" The next wave of competition might be whose moral narrative is more persuasive and compliant with regulations. • For the crypto/Web3 community: Decentralization slogans are loud, but training and alignment remain highly centralized—the value layer is locked up by a few labs and advisory groups, and on-chain governance is inaccessible. • For competitor narratives: Some are already using "We don't hire clergy, we only talk about rationality" as a differentiator (truth aside, traffic is guaranteed). My opinion: It's not shameful to engage in traditional dialogue; what's shameful is pretending to be a "neutral algorithm" without publicly disclosing the source of its preferences. An 84-page charter can be written, but users should at least know: when faced with euthanasia, war, or religious conflicts, which faction's tone the model resembles more. Do you think AI should be taught morality by religious figures? Or should it only accept rules of falsification? Let's argue in the comments section Source: https://(apnews.com)/article/anthropic-moral-artificial-intelligence-religion-2bb91e3334fd6d732edf22288cabf4ec https://www.anthropic.com)/news/widening-conversation-ai https://((x.com))/unusual_whales/status/2116681877788778587 https://((x.com))/i/trending/2116701144328384756