Anthropic and OpenAI attended the Faith-AI Covenant in New York, where they consulted with Sikh, Hindu, Jewish, and Mormon leaders about Claude's values. This move follows a previous meeting with 15 Christian leaders. The goal is to find ethical truths for situations without written rules, recognizing that AI cannot achieve universal principles on its own.
AI seeks moral guidance beyond its training data 🧭
The approach recognizes that language models, trained on vast textual corpora, lack an intrinsic moral compass. By integrating perspectives from traditions such as Sikhism or Judaism, the aim is to create a reference framework for unprecedented dilemmas. This process involves translating theological concepts into algorithmic constraints, a complex step that requires defining terms like dharma or tzedakah in code.
When your AI assistant quotes a guru instead of Wikipedia 🤖
Now, if you ask Claude about an ethical dilemma, it might respond with a Sikh parable instead of a user manual. Imagine asking it for advice on not eating the last slice of pizza and it talks to you about karma. At least, when it fails, it can blame a Mormon or a Hindu. Artificial intelligence is becoming interfaith, all because it didn't know how to say no in time.