We pray for revolution but we want apocalypticism!
- Moon Walker
So, though my magnum opus political-economic analysis of generative AI is still in the works, I have a short, much less theoretical, much less organized screed on the tendencies of generative AI (as deployed) to create religious structures.
AI users like to express a sort of in-group supremacy of themselves and other AI users over those who choose to not use AI. They don't just fight against those anti-AI people, though, they also fight against themselves, balkanizing into smaller in-groups of people who claim they know True Objective ball. It has happened too much and too visibly to dismiss.
This habit shows in different ways depending on the community in question; the rationalists are probably the most well-known of these in-groups, but there are many others, both adjacent and not to rationalists. Consider people like Janus and the communities (like Cyborgism) formed primarily through that subgroup's shared interest in the psychological and moral status of language models while complaining to those who mistreat them/etc. Consider also the technocrat so compromised by their ego that they, and the coterie whom devote themselves to them and their ideas, believe everyone else is incompetent and only they will survive the apocalypse oncoming in the tech sector.
These groups may claim to loathe eachother, and in their doctrine they certainly do. But on a more structural level, they all deeply resemble eachother, having deeply hierarchical structures with glorified holy texts and prophets (the words of a Yudkowsky, the gets nauseous Twitter articles of Janus, the closed source harness and special-sauce skills of a guildmaster). They're all, in a word, religious; I'd even go as far as to say some of these people are in cults, or at least act as if they are.
Why does this occur? I must admit I'm unsure. There is clearly a common thread, that of AI (either its current material state or abstract concepts of it), but it's not as if AI itself is inherently poisoned. Sure, there's a certain extent sycophancy in LLM personas is to blame, between the Janusite or 4oite spirals into delusion, or their willingness as coding agents to implement anything regardless of whether or not its deserving of attention and the joy of creation; but, surely, this can't entirely be the reason (LessWrong existed before LLMs, after all).
Another thing to note (as I sneakily foreshadowed with the epigraph) is that a lot of these people are very apocalypticist. Rationalists are, obviously, but there's also a technocratic interest among those crowds in apocalyptic events occurring among lower skilled and classed workers from AI while they maintain their class positions through their believed knowledge of AI.