Here is some shameless, unfounded speculation copied from my site, Farlops Industries, which is up and running again after a hiatus of 7 months:
A week ago I read a short story by Tad Williams about the
spontaneous emergence of sapience from the Internet. It wasn't really
that good. Arthur
Clarke did it many decades earlier and did it much better.
Anyway, uninformed amateur scientist that I am, I'm skeptical
that consciousness will emerge from our computer networks as long as
they are organized as they are. It may be true, using crude numerical
comparisons of moving parts, that the Internet is at least as complex
as a single vertebrate brain. But this ignores several key
issues.
To explain what I mean, let me pose the following images and
metaphors.
- Suppose we have a tiny clump of cells, just recently
differentiated into neurons, that sit at the top of the notochord of a
developing mammalian embryo. This is where all mammal brains, where all
vertebrate brains, start. It begins here.
- However let's further suppose that the embryo is infected
with strange microscopic parasites which have somehow taken control of
each nerve cell, of each cell in the embryo's body.
- These parasites take over and steer the embryo's
histological development to meet their own goals.
- The parasites dedicate the neurons to performing
tasks that have nothing to do with tissue organization or
organ formation.
- The parasites control how the neurons communicate and
function at all levels. None of this communication or function has
anything to do with the normal histological development of an
embryo.
Do you begin to see my point?
What's happening here is that parasites never allow the embryo
to develop sentience because they are using the cells to do things that
have nothing to do normal embryo development. The neurons aren't really
neurons anymore because they aren't allowed to function like normal
neurons.
Substitute "humans" for "parasites," "computers" for "neurons"
and I think it becomes clear. Computers aren't programmed to function
like
neurons or stem cells. They programmed to function like word
processors, e-mail clients, game machines, bank databases, graphics
editors, web servers
and so on. It doesn't matter that we've hooked them all up into a
network. The
communication between these machines is nothing like the communication
between differentiating stem cells in evolving brain tissue.
This is why consciousness will never emerge from the Internet.
It won't emerge until we completely change the focus of all the
computers on it.
It's my opinion that desktop computers, making rough numerical
comparisons of moving parts, are already as complex as individual
mammalian neurons. Computational neurologists have already written
simulation software that can model individual neurons with reasonable
accuracy and speed that can run on ordinary
desktop workstations. Computational neurologists have now moved on to
more ambitious goals.
IBM's Blue
Brain Project uses some of the most powerful supercomputers
in the world to model something called a cortical column. A neocortical
column is an organized collection of about 80 to 120 neurons all
connected together in a space of one cubic millimeter. The surface of
the neocortex of mammals, the part of the brain that
has all the folds, is composed of hundreds of thousands of cortical
columns.
This should give us some idea what we're up against.
We'd need a few hundred thousand Blue Brain machines are all
communicating as cortical columns would before we can start talking
seriously about the strong
formulation of artificial intelligence, let alone the
spontaneous emergance of consciousness from the the Internet.
Please note, I'm not saying it's impossible. In fact, if you read
through this carefully, you'll see that I'm actually saying that we've
made an astonishing amount of progress. It took blind evolution more
than 3
billion years
to arrive at a mechanism as complex as a cortical column. We've been at
this, what? 500 years? Seems to me we're getting very good, very fast!
Moore's law suggests that it's only a matter of time before
comsumer-grade computers can run something like the Blue Brain
simulator as a low level process. Think of the SETI@Home
program
that analyzes small blocks of radio telescope data. This program runs
as a screen
saver during idle time when you're not using your computer for
something else. Now imagine if the Blue Brain simulator was set to work
in a similar way.
The simulator might hog all your bandwidth as it communicated
with other simulators on the Internet. Other times it would be quiet.
It would all depend on what the simulators were thinking about. Some
simulators would have frequent heavy loads while others would hardly
see use at all.
This still wouldn't be true consciousness though because we've
ignored connecting these synthetic neurons to some sort of body or
senses. The closest analogy I can think of is the brain tissue of an
embryo before birth. There is very little sense data and the body isn't
complete yet.
But at last we'd be getting somewhere. Strong AI would be with
reach. Such a scheme might have a lot to teach us about organic brains.
Cross-posted from Farlops Industries!