To make AI morally good, tech companies are turning to religion, in spite of doubts

As tech companies rush to develop the most advanced artificial intelligence, they are turning to religion for guidance on creating ethical AI models. Chris Olah, co-founder of Anthropic, recently commented on this trend during a presentation at the Vatican's AI encyclical. Olah stated that questions about morality are better addressed by traditions that have grappled with such issues for centuries.

However, the approach taken by these companies to religion has been criticized. Scholars argue that religion cannot be separated from its cultural and political contexts, and thus, using it to train AI models risks prioritizing certain traditions over others. For instance, the Abrahamic religions, which are scriptural in nature, may not align with all faiths that do not see themselves primarily as ethical systems.

Despite these concerns, companies like Anthropic, Meta, Google, Amazon, and Microsoft have engaged with religious leaders and scholars to ensure the development of morally good AI models. These meetings aim to create an "aggregate of humanity's moral view" that can inform the AI's behavior. OpenAI CEO Sam Altman has expressed a similar sentiment, stating that ChatGPT should reflect a weighted average of moral views across humanity.

The meetings between tech companies and religious leaders have been described as rigorous and thoughtful, with participants coming from over 15 religious and cross-cultural groups. However, there are concerns that this approach may inadvertently reinforce a narrow view of religion as a set of rules or principles divorced from its messy, practical aspects.