On 5/27/2018 1:22 AM, smitra wrote:
This is a physical version of what Bruno has been talking about on
this list.
With "strong AI" I mean that any simulation of a person generates the
mind of that person, and the subjective state of that person is
independent of how that simulation is performed. So, what matters is
if certain computations are performed correctly, not on how those
computations are performed.
This is actually ambiguous. First, computation is an execution of a
function: so being "the same function" can only be defined if all
possible inputs are defined; which is excessive. Can the output be the
same only for some small subset of inputs?...for only one? Second, what
does it mean for the same function to be computed differently...does it
mean the same output of the human body given the input of the whole
environment?...or does it mean the same input-output of the amgydala,
the visual cortex, the cerebrum, the cereberal cortex, the frontal lobe,
parietal lobe,.... Exactly where is input boundary and the output
boundary.
I don't think such a boundary can make sense. I think the "mind" only
exists in the context of a world, a world which is effectively physical.
While this doesn't seem to have anything to do with quantum mechanics,
the MWI etc., it actually implies the MWI. This follows from the
assumption that whatever subjective experience is generated by the
simulation is independent of the implementation of the physical device
performing the simulation. The time evolution operator defines a
mapping from the past state of the device to the future state. This
implies that the simulation of a person in a certain state is also a
scrambled version of the simulation of that person at some time in the
future or the past. Since by implementation Independence that
scrambling should not affect the subjective experience, one has to
conclude that the consciousness of the person at any time in the
future and the past is also generated by the simulation.
If we start running a simulation at t = 0 s and the simulation is
programmed to shut down at t = 1000 s, then at any time, the running
of the simulation generates the consciousness of the simulated person
at all times between t = 0 and t = 1000. The computer plus the
environment it is in contains all the information about the past and
present, including how the computer was powered on, the simulation was
started and how the simulation will end etc. This then contains all
the information about the person between t = 0 and t = 1000, but not
before or after this time period.
The next step is to consider running a simulation that depends on the
outcome of a spin measurement in the real world. We polarize a spin
the positive x-direction and measure the z-component, feed the result
of the measurement in the simulation and that then affects the
simulation, the simulated person will be made aware of something that
is different depending on the outcome of the measurement.
The moment this experiment has been set up and is ready to go, the
time evolution of the entire system that includes the experimental set
up., the computer and everything else that is of relevance, is fixed.
But this time evolution will bring the system into a superposition of
the two experimental outcomes of the measurements and its consequences
for the simulation.
This implies a completely deterministic evolution. In that case it is
not clear how the Born rule is implemented in the simulation.
It then seems to be a matter of belief in the MWI whether or not one
should believe if both possibilities happen or if only one of them is
going to be real. But if we accept strong AI, then we have to accept
that the physical state of the system before the measurement is
performed, will also generated the consciousness of the person after
the measurement. Because the information present in the physical state
before the measurement contains both branches, it generates the
consciousness of the two versions of the person in both branches.
But of course the simulation can simply stop all the branches but one.
Brent
Finally, one can argue that the conclusion applies to real persons
that are not simulated by computers, because the brain computes us,
and in the above argument it doesn't matter if you use a computer or a
biological brain. In fact one needs to appeal to the entire
environment containing the computer, and to get to a rigorous argument
this has to be taken as large as the size of the lightcone starting
from the moment the simulation starts.
Saibal
--
You received this message because you are subscribed to the Google Groups
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to everything-list+unsubscr...@googlegroups.com.
To post to this group, send email to everything-list@googlegroups.com.
Visit this group at https://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/d/optout.
