Raw LLM Responses

The "Superintelligence" crowd is obsessed with the idea that an artificially intelligent machine would have the capacity to design a more intelligent artificial intelligence, and if the resources were available this series of increasingly intelligent machines would eventually lead to a superintelligent entity that would be beyond human comprehension. If we are told that a [hypothetically] superintelligent machine would be 1000 times as intelligent as any human, on what terms would we be making such a comparison? Does intelligence have a unit measure, so that this entity over here can be said to be twice as intelligent - or 1000 times as intelligent - as that entity over there?