[FoRK] Brain mapping and the connectome
sdw at lig.net
Tue Nov 10 10:19:08 PST 2009
Jeff Bone wrote:
> Ken writes:
>> if it's not essentially "us" I'm not interested.
> Me either: problem is, I don't think it's *possible* to have a
> non-trivial, objective, useful classification boundary between "us"
> and "them." And, in fact, I think it's pretty close to provable (cf.
> Ship of Theseus-type arguments.)
Agreed. We're already augmented, and we're going to become far more
augmented. There are many examples of extension of our mental bodies
through tools and out into the world in other ways. This includes new
senses, adding a compass sensor for instance.  Whether those
extensions are integrated or even replace some of our bodies is just
>> And I haven't seen Consciousness
> Capitalized, no less!
Consciousness is not that big a deal. Really. Interesting, and
important to understand, but it is really just a matter of self-aware
attention tracking. If what is really meant is having enough cognition
to be self-aware in a meaningful and useful sense, then that is closer
to the interesting problems of AGI. The enabling of general cognition
to support consciousness is hard, consciousness itself isn't so impressive.
> I thought you said you weren't a theist (or were anti-theist, or what
> have you.) Deus ex machina, much?
atheist->Atheist->anti-theist? Is an anti-theist a proselytizing
Atheist? Seems like a matter of indifference, degree, and politeness.
>> Anything I read about AI
> Apparently either isn't much or isn't the right stuff. ;-) We
> *clearly* have learning; what we don't yet have is generalized
> agency, situation, context, etc.
We are far far closer to usable theories than 10 years ago. We have a
lot of things working well, providing a lot of useful sensory handling
for instance: speech recognition, vision of all kinds including almost
instant accurate 3D models of everything visible in video. Available in
open source software...
The useful thing about mapping better how the brain works is to learn
new mechanisms, structures, methods, and generate creative ideas to help
us solve problems better. We don't need to be able to exactly map a
running system to learn a lot of useful stuff from it.
If we were to find some inscrutable alien computer technology, even
knowing what molecules, energy, frequency (or not), they were using
would give us extremely useful clues to areas we hadn't thought were
useful or competitive.
Once you understand the principles involved, insisting that you have to
know the quantum state of every subatomic particle involved is obvious
overkill. Knowing and trying to duplicate the exact state of a running
system might require it, however that is not really a goal unless you
are building a Trek transporter.
We keep getting stuck in local minima and we need constant ideas and
evidence to climb out. Nature / evolution has done pretty well, but it
has plenty of local design minima too. The "mulefa" in the His Dark
Materials / Amber Spyglass novel was a brilliant solution to "nature
didn't invent wheels". 
>> It still sounds to me that we're quite a ways away
> It's worth nothing that whatever it sounds like to you, the various
> subject matter experts quoted in various of the links in this rambling
> pseudo-thread seem to have a different opinion. (Others disagree.
> YMMV. My money's on the connectionists --- *very* literally.)
I'm glad you find trading interesting. I can't get into it. I have a
usefulness filter that just won't quit.
Now, if you are discovering useful techniques and publishing them,
great. Seems like anything useful is black-holed in that realm though.
> How far we are depends on how you measure distance. We clearly have
> *nothing* even remotely approximating the order of complexity of a
> human neocortex. However, that doesn't mean that such is untenable in
> relatively short human-subjective timeframes (i.e., if accelerating
> change laws hold, for example... Consider again the lifecycle of the
> Human Genome Project. We're recycling arguments here, except we
> aren't arguing --- you're just reasserting.)
I expect non-linearity. We've seen it before. We couldn't fly for a
long time, then we could, and a short time later we're walking on the
moon. We barely knew what to do with electricity, barely had radio,
then suddenly we had lights, TV, cell phones, etc. We have evolved our
evolving so that new discoveries don't take 150 or 60 or 20 years, they
get to market in 18 months or less many times, sometimes creating whole
new markets in a few years. AI will be the same.
More information about the FoRK