The phrase “inaccurate simulation” led me to expect discussion of the possibility of a nonmaterialist nonagentic substrate—a simulation where errors look less like random bitflips and more like AI confabulations. Random, non-goal-oriented, semantically meaningful deviations from a purely mechanistic universe, like the result of an LLM (maybe from 2023 or so) fed a description of history and asked to extrapolate the next decade.
The phrase “inaccurate simulation” led me to expect discussion of the possibility of a nonmaterialist nonagentic substrate—a simulation where errors look less like random bitflips and more like AI confabulations. Random, non-goal-oriented, semantically meaningful deviations from a purely mechanistic universe, like the result of an LLM (maybe from 2023 or so) fed a description of history and asked to extrapolate the next decade.