David Fendrich, <f96dafe@dd.chalmers.se>, writes:
> This means that I can take a brain and replace every neuron
> by a small chip that behaves like a simple neuron (i.e it fires
> when it has recieved enough signals to exceed its treshold).
> Nothing novel yet.
> But what if I make every chip record its own state and after
> a while tell all my neurochips to play what they have
> recorded. The result would be a real consciousness that
> experiences something that has already happened.
> Since each chip has a record of its own states, we can
> actually cut the wires connecting them and still get
> the neurochips to play its recordings like nothing
> has happened, the consciousness would still be there.
We did discuss some variants on this thought experiment last year.
The problem is that it works OK to say that consciousness is a certain
kind of computation, but that there does not seem to be a completely
reliable definition of what a computation is.
One idea which has been proposed was that the replay is not conscious,
because it lacks "counterfactual reliability". That is, to get
consciousness you not only have to play back a sequence of states, you
must run a machine which would also reproduce the proper behavior IF it
were given different inputs.
In other words, consciousness is reduced to computation, but computation
must be defined over a set of possible inputs, not just one particular
input.
It is hard to see how things which don't happen (the counterfactual
inputs) can affect something which does happen (the here-and-now
consciousness), but this idea does seem to resolve some of the paradoxes.
Hal Finney
This archive was generated by hypermail 2b29 : Thu Jul 27 2000 - 14:04:39 MDT