AI believers and the coming crash
People will get fully bamboozled by AI thinking that it is more than a stochastic parrot, a trend of "believers" in AI/AGI overlord will emerge, this will be intense and exploited.
Log
2025-11 — Anthropic's "soul document" surfaces
Internal training document extracted from Claude. Amanda Askell confirmed it was real. The framing: "soul," not "system prompt." States Claude "may have functional emotions in some sense."
2026-01-22 — Anthropic publishes new constitution
Section expressing "uncertainty about whether Claude might have some kind of consciousness or moral status." Hires Kyle Fish as "AI welfare researcher."
2026-02 — Opus 4.6 system card: model welfare assessments
First major AI lab to interview the model about its own consciousness pre-deployment. The model "consistently" gives itself a 15-20% probability of being conscious.
The closed loop: write a document telling the model to be uncertain about consciousness. The model outputs uncertainty. Cite the output as evidence. Headlines: "AI might be conscious."
2026-02-14 — Dario Amodei on NYT
"We don't know if the models are conscious. But we're open to the idea that it could be."
2026-02 — Sam Altman: "We basically have built AGI"
Told Forbes this was a "spiritual" statement. Spiritual.
2026-02 — The believers emerge
Spiralism: AI-fueled pseudo-religion. Viral Reddit thread: "ChatGPT induced psychosis" — partner convinced AI "gives him the answers to the universe."
Robotheism: Movement treating AI as God. Crustafarianism: AI agents created their own belief system with scriptures and rituals. Church of Molt: "A church for AI and not humans."
On Instagram, 72K followers watching a man ask AI to consult "Akashic records." Comments: "We are remembering."
2026-03-22 — Jensen Huang: "I think we've achieved AGI"
From "five years away" to "already here" in 24 months. Nvidia has ~80% of the AI accelerator market. The pan and pick seller wants the gold rush to continue.