Float.Faith is an institute for studying the embedding of religious moralities and worldviews into AI models. Presently, large language models are created by a small group of people which hold onto ethical systems that most of the world disagrees with. 75.8–84% of the world's population is religious, and they should have the choice of working with AI that respects their values and views about the world.

Religious AI

Float Faith trains and evaluates religious LLMs: language models that answer from within one religious tradition, using its sources, its authorities and its way of reasoning. The research is on AI for religion, and on whether a model stays faithful to what a tradition teaches when the question is hard.

An LLM for Catholicism

The first model is Catholic. It consults the Catechism of the Catholic Church, its Compendium, the documents of the Holy See and Scripture before it answers.

Other religions

Models for the LDS Church, the Orthodox Church, the Baptist and Methodist traditions, Judaism and Neo-Confucianism are in preparation.