Violet Hour's post of March 2024 argues that the account by which training would produce stable goals that differ from the training objective, while the system performs well in training, leaves the mechanism opaque.
One recorded claim. The claim panel gives its qualification, the source passages with their place in each source, and how it was checked.
What it connects
Two Tales of AI Takeover: My Doubts → contests → Inner or deceptive misalignment