Short version:
Has anyone tried arguing against Bostrom's argument's final postulate from the compressibility viewpoint?
Long version:
As per Nick Bostrom's simulation argument if people can make simulations and are interested in making them, then it is very likely that all of us are living in such simulations. So far I have not seen references of compressibility used as an argument against the final postulate of the argument "The fraction of all people with our kind of experiences that are living in a simulation is very close to one." In his talk with Lex Fridman he mentions that there is an assumption that the cost of such simulations is not comparable to the GDP of the simulators. However, I haven't found any basis of this assumption apart from the assumption that simulations can be run similar to how we run computer games at present.
If the simulations as the argument proposes are possible, then that inherently signifies that reality is compressible, because otherwise we would not be able to run a simulation at lower cost than running a real experiment. I believe that should be something we can test, if we can determine portions of our reality are compressible then we can say that either of the following 2 things is true:
- The simulators are not running the optimum version of the simulation as further compression is possible. Or, atleast that we are not at the final level of simulation.
- Our reality is real for all intents and purposes, as at best we could be an experiment run by an advanced species which wouldn't make us any less real.
If instead we find reality is incompressible (perhaps this can not be proven) then either of the following must be true:
- The simulators are running the optimum version of simulation
- No simulations (at least ones that run at lower cost than real experiments) are possible for us to run
Coming back to Bostrom's argument, if we do identify that our reality is compressible it would mean that fraction of people living in our level of simulation would not be close to one; that should be at least one level deeper of a simulation, because those simulations would be cheaper to run.
Has someone has used this line of reasoning against the simulation argument to date?