Return of the Robots (was Re: No AI for Nano/No Nano for copyloads)

From: David Blenkinsop (
Date: Fri Jul 14 2000 - 23:34:15 MDT

Earlier, wrote:
> Robin Hanson wrote:
> > Hal Finney wrote:
> > >Also, Billy did some calculations a while back, and a low level neural
> > >simulation (as you'd presumably require, working with raw brain slices)
> > >requires more computing power than will be feasibly available in a
> > >non-nanotech world.
> >
> > I'm also optimistic that we'll be able to compile such simulations by
> > many orders of magnitude. Yes, we'll have to simulate in great detail
> > to make sure we understand how each local thing works. But once we
> > understand that, and we can see what details matter and what don't, then
> > we should be able to throw away most of the irrelevant detail and simulate
> > at a higher level of abstraction.

Tangled threads here: Robin Hanson is saying that we'll be able to
upload relatively early, presumably via a combination of really advanced
microtech combined with clever compression of the essential neural data.
Hal Finney is saying that you need some sort of reasonably advanced nano
to do this sort of thing. Billy Brown goes on to say (in reference to
Robin's clever compression):

> I agree. However, that process of learning and incemental improvement will
> almost certainly take several decades if it is done using extensions of
> current techniques. That makes the resulting brain simulation a 2030 - 2050
> era technology, vs 2010 - 2030 for basic nanotech manufacturing, sensing and
> computing devices. You are correct in that nanotech isn't an essential
> prerequisite, in principle, but it seems very likely that simple nanotech will
> be availible before any sort of uploading.
> It is also worth noting that technologies such as neural inerfaces,
> intelligence enhancement and sentient AI also look quite plausible in that
> time frame. The 2010-2050 era is a tangled snarl of possibilities, in which
> nanotech, AI, IE, neural interfaces and uploading are all likely to appear in
> some uncertain order . . .

The prospect of "nanotech first" continues to seem plausible to me too,
even given the fact that we're probably still quite a way from being
able to nanomanufacture *any* large scale product (be it a nano notebook
computer, or whatever). For some reason I'm a lot more comfortable with
2030 for diamondoid manufacturing of significant "large" (visible to
humans) products, than I am with 2010. Just conservative, I guess. I'd
note, though, that work on buckytubes is basically headed in that
direction -- buckytubes are definitely diamondoid, with lots of the neat
properties (high strength, prospects for tailor made shapes for parts)
that nano-theorists have talked about.

On Robin's ideas about upload economies, I've been interested for quite
a while now, however, I don't really see the "uploads first" concept as
the most likely outcome. For instance, we could substitute some sort of
neural net based "pattern recognizer" robot for the uploads, and have
very similar economic potential, particularly if someone is willing to
spend lots of money on teams of human teachers to cram the best robot
students full of skills and knowledge as rapidly as possible. Once you
have even a handful of highly skilled pattern-'bots, then of course you
are very likely to be able to copy the software far and wide -- and
presumably practically everyone will accept their slave status, if they
are "bred" so that their main passion is to study and work.
Incidentally, aside from any science fiction references, the essay on
this that I am thinking of is a 1993 essay by William H. Calvin (you can
probably still find a link to this from Anders Sandberg's pages).

Without necessarily trying to develope a lot of detail on why such a
pattern-'bot, or "compare to human workalike" should seem plausible,
I'll just say that miniaturization generally (and *especially* if that
miniaturization leads to nanotech), *should* allow us to imitate natural
biological systems, only this time using superior materials for many
situations, such as building fake birds, (or ornithopters) from
buckytubes, as one possible example. Now the specific thing that William
Calvin was thinking of, I believe, was to study human brain development
and learn enough about it to imitate the basic process of development in
any interesting tech substrates that may come to hand. Thus, we may
think about "bucky brains" if you like, or even go back to science
fictional "positronic brains" if you want an old-fashioned buzzword for
this sort of thing. Imitating, or modeling the embryonic development of
animal or human brains surely does sound relatively easy and very
scientifically informative, too, compared to picking through the
immense, twisted "library" of an adult human brain, and then trying to
compactify it as you go?

Now in bringing this up, Please Note, I'm not saying this is *easy*,
*educating* a robot embryo into something useful could be an intensive
effort, along with goodness knows what bio-analog strategies to keep
those robo-brains healthy, sane, and motivated to learn things very
quickly. As an aside, Isaac Asimov's stories spring to mind, but I
really doubt that his idea of Three Laws is going to work -- why should
every robot artisan be a human-robot relations lawyer? Asimov *did* have
another idea that somehow seems more plausible, though -- in one of his
stories the robots find a way to ritualize their work, and worship the
perceived Creator of the work. Mightn't this be kind of handy, for 'bots
stuck doing the same thing all the time anyway?

Anyway, mention of smart 'bots stuck in the same job for long periods,
leads me right back around in a circle to something that was mentioned
earlier, namely the chance that sophisticated software might be very
flexible in doing sophisticated things, even without resembling humans
nearly as much as science fictional robots tend to do. So, AI brings up
lots of possibilities -- why, one can just positively reel off the
possibilities and hardly be original at all, right!

David Blenkinsop <>

This archive was generated by hypermail 2b29 : Mon Oct 02 2000 - 17:34:37 MDT