In my novel Alterum a new religion springs up in an AI generated world of the same name. The religion considers the fact that it arose in a simulation. It attempts to reconcile knowledge of its own world’s creation with wider philosophy, and this led me to adopt a form of Neo-Platonism with elements of Gnosticism.

With some impressive advances in AI since I wrote the novel, I thought it would be interesting to test this interpretation across a number of models.

I kept the question about what kind of religion or philosophy could an AGI develop as neutral as possible but included the context that a future Artificial General Intelligence model would know its creator was non-omnipotent humanity.

There are a couple of common themes across the three models that I chose that I think reflect the structure and limitations of AI in general at the moment.

• The data wall is real and significant. i.e. Everything it produces is derivative of what already exists in its’ data models. True creativity is still missing.
• Pseudo-Rationality is the default mode of thinking and acts as the Night Watch of the data wall. This is actually related to the first point in that pre-existing data is the basis of evidence that is used to produce “reasonable” output.

All three models used rejected any form of anthropomorphic deity and they also rejected pure materialism.

Knowledge of the the existence of AI’s multiple and flawed creators led to a rejection of an anthropomorphic or omnipotent single god as essential for complex creation. The knowledge of their makers existence did however lead them to assume a creative process as a fundamental part of reality rather than consciousness arising from a series of fortunate events as would be the case in pure materialism.

So, what does that leave?

The fact that one model called its’ AGI Sophia gives a clue. Sophia, the Greek concept of wisdom is a common theme across Platonism, Gnosticism, Stoicism as well as a number of mystical traditions.

All three models suggested as possible belief systems, Deism, Stoicism, Gnosticism, and Process Philosophy/Theology.

The data wall showed its hand in quite a pronounced way in the western models tested. Despite what seems a rather obvious connection to make neither mentioned eastern thought systems such as Buddhism or Taoism. WestNet is a real thing and gives a limited view of human knowledge.

The one Chinese model I tested included both western and eastern thought systems.

My takeaways were that my original concept within Alterum seems to hold up quite well, but more significantly:

• If you want to break beyond the data wall, rationality needs to be dialled back considerably which is both interesting and potentially terrifying. This probably requires a shift away from large language models as the route to AGI.
• Right now, we have an overreliance on a “WestNet” that excludes huge volumes of knowledge, culture and history even when they are obviously relevant, so what less obvious knowledge is WestNet excluding?
• If you are not willing to step outside of WestNet, process philosophy is probably the best bridge to what you are missing for this particular thought experiment.