Right. We apply some of the best minds and most powerful compute resources in the world to predicting the weather and yet can't get it right within a reasonable degree for much more than a week. For an AI to try predict what humanity will do to stop it will surely look a lot like paranoia.
> Switching the machine off won’t work for the simple reason that a superintelligent entity will already have thought of that possibility and taken steps to prevent it.
Also the statement presumes that self preservation is the super intelligent thing to do.
Also the statement presumes the superintelligent entity will be capable of taking all possible steps to prevent this. Imagining a machine capable of such steps quickly devolves into imagining a machine already in total control of the planet, which is begging the question.