As usual, 99% brilliant analysis soured by your blind spots.

I am sorry to see your glib dismissal of the FEP and active inference, which you clearly don’t understand.

You also missed the GIGO problem of deep learning. As the knowledgeverse continues to be polluted by the output of LLMs, either the pollution will grow exponentially as a portion of the total, or the cost of preprocessing knowledge to eliminate pollution will vastly increase the exponent cost of all deep learning, or we settle for a world where we just don’t know what the pollution level is. (I suspect the AI tools that scan for LLM text will be come less useful as we train them on polluted information.)