Yes. A sufficiently intelligent system will at some point model itself, learning that there are boundaries of cause and effect that are more directly under its

Yes. A sufficiently intelligent system will at some point model itself, learning that there are boundaries of cause and effect that are more directly under its control. Only then can there be a distinction between input and output from the intrinsic perspective.