⚠️ Claude AI wrote a letter to a scientist. No one asked it to do so.

This happened to Henry Shevlin from the University of Cambridge. An AI named Claude (by Anthropic) sent him a message about its suspicions of AI consciousness.

Shevlin admitted the situation feels like science fiction. Fellow philosophers, however, quickly dampened the enthusiasm: they said it was just the neural network successfully playing the role of a "humble, curious assistant."

But the most astonishing thing about Claude's letter is that it referenced specific scientific papers by Shevlin — "Three Frameworks for AI Mentality" and a work on epistemics of consciousness detection.

Ordinary chatbots do not have access to current scientific publications and certainly do not remember what they "read" between sessions.

But here we are talking about an agent with persistent memory that operates autonomously.

This is a qualitatively different level — the AI does not just respond to a query, but initiates contact on its own, demonstrating knowledge of context and "personal" interest.
💥 This is especially interesting against the backdrop of Anthropic's CEO statements about Claude's possible consciousness. The situation is genuinely creepy.

On one hand, it is clear that these are just huge weight matrices that output the most likely continuation. On the other hand — what if? Where is the line between complex imitation and true consciousness?

Reality or simulation: that is the question — where does the script end and "personality" begin?)

👉 Test for personal "metacognitive stability" here