From: Dan Fabulich (daniel.fabulich@yale.edu)
Date: Tue Jul 11 2000 - 16:58:21 MDT
Summary: If interesting nanotech is a long way off, or requires AI to
work, then any immortality race against AI will fail, especially if
seed AI is coming soon.
Recently, my thoughts returned to Lyle Burkhead's Geniebusters
website, at www.geniebusters.org.
I've always regarded his claims to be wrong, not the least in part
because paying an AI is very different from paying an economy full of
hungry mouths, and because, once we do figure out how to program
nanobots to build a battleship, we should be able to make battleships
on the cheap.
But recently a claim that he was making sunk in. He asks:
"Can you think of any system (of any size, localized or scattered)
that can take anything apart, while recording what it removes layer by
layer?"
This question is supposed to lead us to the conclusion that, no, we
can't really think of anything like that, and that, therefore,
disassemblers are a long way away.
Of course, this is nonsense. We don't disassemble buildings like that
because it's simply cheaper and easier to consult previously made
blueprints to analyze buildings. But where we don't have blueprints,
disassembly is exactly what we do, though the process has yet to be
automated.
But there's something deeper to this, if you look at it.
Right now we already have the tools to build buildings manually. Yet
we don't yet build buildings by dropping a pile of metal and concrete
near the area where we want a building built and then instructing the
robots to "make me a skyscraper, according to these blueprints."
(Compare this to Drexler's picture of building a diamondoid rocket
engine armed only with assemblers and the raw carbon required to build
it.)
Though we already have the manipulators, we don't have robots that can
build buildings. Nor do we have robots that can build other robots,
at any size, though, again, we already have the manipulators,
sometimes even computer-controlled manipulators, to do so.
For a long time I was of the belief that the advent of the first
computer-controlled universal assembler would trigger a huge growth
market in designs for things to be built by nano-assembly. Now, I'm
beginning to doubt it. Because once we get the universal assembler,
we'll be in exactly the same place we are now with conventional robots
and buildings.
Until we can automate the construction of a building using
conventional robotics, we should not expect to be able to automate the
construction of any interesting large structure by nano-assembly.
Will nanotech change this playing field in any significant way? Well,
somewhat. Maybe the reason we have yet to automate the construction
of buildings is because the process is difficult, and because there
wouldn't be THAT much payoff in the investment. With the coming of
universal assemblers, the possibility of the object copier, which can
disassemble anything, and rebuild as many of the object as you like,
will seem feasible. You might imagine that there would be a VC race
to see who could fund the first one of these, since the payoff would
be so large. (But would it? How are you going to get a return on
your investment if anything boxes can also make anything boxes?)
But even then, while we might expect more money to be thrown at the
problem, it'll be a huge and extremely difficult problem to solve. We
might not expect to solve it this century, and to the extent that we
do solve it, the solutions will move at the pace of robotics today:
unstoppably, but gradually.
Of course, an AI could solve the problem; arm a transhuman AI with
nanotech and step back. (Or, better, get in a rocket and fly away;
get as far out of its way as you can and hope the AI will be friendly.)
---
So, to the best of my knowledge, there are only a few ways that we
know of to get us immortality: we might use nanotech to keep ourselves
in good shape, we might try uploading into a computer, or we might
develop an AI to figure out how to do one of the above. In the past,
some have suggested that these technologies are in a race, and that
something bad might happen if one of them happens before the other.
(... If nanotech develops significantly before AI or uploading, we'll
kill ourselves with grey goo... if AI happens before uploading, it'll
kill us all ... if uploading happens before AI, the first guy to
upload will kill us all...)
But if nanotech-without-AI is a long way off, and AI and nanotech are
both coming soon, this changes the playing field drastically. Now, we
don't need to worry about developing nanotech before AI, because AI is
the enabling technology for nanotech! Similarly, it makes uploading
as an alternative to AI development look silly, because nanotech is
almost certainly an enabling technology for uploading, (since we can't
reasonably expect to upload until we get reasonable disassemblers to
work on our brains,) and AI is required for nanotech.
In short, you can't win a race against AI, because you have to use AI
to cross the finish line.
And, of course, if and when a seed AI crosses the finish line... well,
all bets are off.
-Dan
-unless you love someone-
-nothing else makes any sense-
e.e. cummings
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:29:52 MST