Robin Hanson wrote:
> I don't think nature is very well compiled as it is. Nature happened
> upon neurons
> as a computational mechanism and used it, even though most of
> what happens in a neuron is irrelevant to the computation.
I agree. Buy I expect we will have to unravel all of the mind-boggling,
inefficient complexity of the existing system before we understand its
behavior well enough to make a more efficient replacement that accurately
duplicates its behavior. Otherwise we'll end up with inaccurate simulations
that don't work.
> I'm not sure we know much about how nanotech changes the relative rates of
> progress.
Not in detail, no. But that doesn't mean we can just dismiss the subject,
because it is obvious that there is a lot of potential for significan
tinteractions. Until we have a better picture of those interactions (which, of
course, means a lot of difficult research), it is going to be very hard to
assess the plausibility of any particular technology scenario.
> I don't think you understand the economic argument here. Even ignoring
speed
> advantages of uploads, since it is cheap to create uploads, the supply
> increases quickly, which lowers the market wage. I did try to explain this
> stuff at http://hanson.gmu.edu/uploads.html
That URL doesn't seem to be working, but I beleive I remember the paper. As I
recall, you analyzed the effects of rapidly adding uploads to the labor pool
in considerable detail. However, your analysis focused exclusively on the
supply side of the equation - there is no consideration of the fact that the
uploads also contribute to agregate demand, which would tend to offset the
downward pressure on wages.
I also think that your 'fast, cheap copying' assumption is an artefact caused
by projecting a very advanced technology into a world that is otherwise
unchanged. If uploads actually existed today, copying one would be a
multi-million dollar operation involving weeks or months of work by a team of
highly-paid specialists. A future where uploads can be created and duplicated
like text files is a future where virtually everything important is software,
and therefore capable of equally rapid duplication. It is also a world right
on the verge of full-fledged mental modification technology, which takes us
beyond the realm of easy prediction.
> Nine years of AI research left me with definite impressions. And we already
have
> very cheap animal intelligence - real animals are cheap, and they come with
bodies
> too. But very few organizations manage to replace much human labor with
animal
> labor.
Considering how limited our ability to train animals is (compared, say, to our
ability to 'train' computers), that is hardly surprising. Robots are a rather
different case.
Billy Brown
This archive was generated by hypermail 2b29 : Mon Oct 02 2000 - 17:34:38 MDT