Re: Uploading, info theory, and threads of consciousness
Thu, 31 Oct 1996 17:04:11 +0000

> >The problem is that people are viewing information as though it were
> >physical matter. Information can never be moved, it can only be copied.
> >Uploading is the transfer of information. The inability to move
> >information (as opposed to copying) is a fundamental limitation of all
> >information systems.
> >It appears that uploading should not be possible for these reasons. The
> >best we might hope for would be to augment our existing systems.
> We can still live in a virtual world though with the laws of physics
> altered at our whims and AI civilizations created and destroyed for your
> entertainment. (Oh no! Here comes the AI civil liberties organization!)
> In this scenario we do have the choice of either being a brain in a jar
> with i/o streams feeding into a PC or we can augment our existing body
> everyway we can think of and have a brain-PC interface builtin to our
> genome such as communication with a PC via bat radar signals or tendrils on
> the back of the neck which are a direct link to our brain. If this is true,
> it doesn't matter much to me because I'd be satisfied with the 'greek god'
> idea because I could still create entire universes in a virtual world as
> well as have access to the continuously growing pool of human intelligence
> plus gain immortality. Can silicon actually support life? If this is true
> that we are required to 'move' data then Strong AI will never come about or
> live in a quasi-consciousness where it continually lives/dies/lives/dies.
> John, I'm looking for a reply on this one.
Being able to "jack in" to cyberspace is only marginally useful if
you don't have the processing power to go with it. Sure, you may
have access to vast quantities of information, but if you are still
processing it or thinking about it at tissue speeds then you aren't
much better off. It would amount to really nice visualization
hardware/software. You might be able to compensate by having AIs
preprocess the information, essentially offloading some of the
mundane thinking required, though.
It might be possible for Strong AI to develop, but I don't think
it would likely develop spontaneously on a dynamic network. Part of this
might just be a limitation in our thinking. It is hard to build a
stable information architecture or operating software on a system
where your resource structure and capabilities is constantly
changing. Unfortunately, this is pretty much what the Internet looks
like. Sure, it is possible to build a distributed system where the
software is running in pieces concurrently among a number of systems.
You even get some level of fault tolerance this way. But when
building any type of really sophisticated distributed system, you
usually have a very concrete idea of the structure and capabilities
of the overall system. As I said, it is probably possible for a
Strong AI to develop on a dynamic network, but it is both beyond our
capabilities to model this type of system and quite possibly beyond
the capabilities of the operating systems currently running on the
much of the network to coordinate in this fashion. Bandwidth and
latency is also an issue. A distributed intelligence would require
some level of synchronization between its distributed components to
operate in any type of coherent fashion.
Imagine if our neurons had to communicate with each other via UUCP!
The activity of a distributed network AI would probably look
something like an ant colony. Individual components are kind of
stupid, but the components as a whole show considerable
intelligence and direction.
I think a more likely scenario would be an AI being developed on
a structured secure network, and expanding into the internet from
that point. This would give the AI a secure static base from which
to coordinate its consciousness. It would also give its core entity
a much higher level of data and process integrity.

The problem with the human brain is that we are more or less
locked into the architecture that biology gave us. Even if we
replaced every neuron with optical switches, we would still be
limited by the architecture. It would be like running a '286
computer at 500MHz. Fast, but very limited.

-James Rogers