I guess overuse and epistemic capture are kinda descriptive of mania, but I think they’re consequences of something else. (I don’t think you were trying to make the ultimate comprehensive true name classification of course.) I think getting more sense of the something else might help people handle it better. In the case of AI mania, it might be something involving “I suddenly expect all my intentions / ideas / plans to be feasible, which makes me extremely optimistic about everything and/or feeling very high opportunity cost about everything so I’m compelled to keep using the thing and also all my ideas are popping themselves up more strongly because they are excited that they could be easily implemented and that’s abnormal / too much ideas popping up”.
It also feels inseparably related to the shape of the “not AGI” moments—LLM’s somehow not being able to grasp writing style, taste, and intention, hitting manual implementation guardrails on something like “build a $1M business please” and the contrast of being able to quickly make limited-usecase tools and the gaps of how many aspects of serious work are still not yet automated.
I guess overuse and epistemic capture are kinda descriptive of mania, but I think they’re consequences of something else. (I don’t think you were trying to make the ultimate comprehensive true name classification of course.) I think getting more sense of the something else might help people handle it better. In the case of AI mania, it might be something involving “I suddenly expect all my intentions / ideas / plans to be feasible, which makes me extremely optimistic about everything and/or feeling very high opportunity cost about everything so I’m compelled to keep using the thing and also all my ideas are popping themselves up more strongly because they are excited that they could be easily implemented and that’s abnormal / too much ideas popping up”.
It also feels inseparably related to the shape of the “not AGI” moments—LLM’s somehow not being able to grasp writing style, taste, and intention, hitting manual implementation guardrails on something like “build a $1M business please” and the contrast of being able to quickly make limited-usecase tools and the gaps of how many aspects of serious work are still not yet automated.