On Wed, Jan 28, 1998 at 02:26:23PM -0800, Hal Finney wrote: > I am still confused about the mapping between TM output and a universe. > A TM produces abstract output, a list of symbols. Do we need to also > specify some kind of mapping between the output and the universe? > > Suppose we have a physical universe with some characteristics, and a TM > which is offered as an implementation for the universe. We want to > understand whether the TM actually does implement the universe. Maybe > this is in Wei's model where you input the coordinates of a region and > the TM produces the state of the region, or perhaps the earlier suggestion > that the TM produces a sequence of strings which represent time slices > of the universe.
I take the position that what we ultimately care about is the relationship between the TM and the inside view (i.e. what we perceive). It may be that mapping the output of the TM onto a "physical universe" helps in this process, but I don't think this is essential. So, the real mapping problem is how to take a string of 0's and 1's and interpret that as a self-aware being and figure out what it is perceiving. I agree that there is a danger in putting too much information/complexity into the mapping/interpretation process so that it is doing all of the work instead of the actual universe. The mapping problem may be a hard problem, but perhaps for now it would be sufficient to assume that only simple interpretations will be used, and that if it looks like the interpretation process is doing too much work, we'll work with a new universe TM that internalizes most of the interpretation process.