macleod@drivax.UUCP (MacLeod) (08/29/88)
In article <572@unisv.UUCP> vanpelt@unisv.UUCP (Mike Van Pelt) writes:
:...The more you look at it, the problem of making a
:machine that can do even what a cockroach does is not as easy as
:it may appear at first glance. (Which is one of the brick walls
:I think the nanotechnology folks are going to run into.)
This may just be a loose analogy, but I have to add that cockroaches are
generalized creatures; even if the reductionists were correct, a description
of their state machinery would probably come out to some nearly-infinite
model that approached a complete mapping of all chemical reactions.
The kinds of machines proposed by Drexler, as I read _Engines of Creation_,
are >specialized< devices that are produced to address specific problems,
like the machine that goes into cut out the Tay-Sachs gene in the ovaries.
Later there will be a *more* generalized "doctrobe" that repairs a host
of problems, but the problem is one of linear extensions to a program, not
generalizing the behavior of something high up on the evolutionary scale.
If problems in nanotechnology arise anywhere, I suspect it will be in
implementation of ironclad error-correction algorithms. I have not
kept up with developments in either subbranch of CS theory, but I
believe that Error Correction is doing far better than Artificial
Intelligence.
Michael Sloan MacLeod (amdahl!drivax!macleod)