[extropy-chat] Futures Past

[Mike Lorrey]

> On 10/10/05, J. Andrew Rogers <andrew at ceruleansystems.com> wrote:
> >
> > The very notion of the Singularity as most transhumanists conceive
> > it is, as best as I can tell, premised on the same poor reasoning
> > that Harvey is pointing out.
> >
> > The Singularity as most people conceive of it essentially requires
> > a hard take-off(*) of some type

I've never thought so. Unlike some here, I've always felt that it would
be a slow swell, primarily because of my practical experience in
business with budget cycles, development cycles, reporting periods,
etc.: the nitty gritty of moving physical objects around in the
manufacturing process from raw material, to part, to subassembly,
assembly, to finished product, to warehouse, to distributor, to
retailer, to consumer, as well as all of the market analyses,
engineering studies, product definitions, and all of the other
paperwork and meetings and decisions and buy-in that precedes even the
start of production.

The fact is that as soon as chip designers were able to design on a CAD
system of any quality, their doubling period should have started
contracting significantly, at least 50% each chip generation. They got
slowed down by other factors in business that limit its ability to
really compress doubling periods, some of which I've mentioned above.

For example, the production and distribution cycle I described above
frequently results in chip engineers designing the newest chips not on
the previous generation of chips, but two generations older, so their
maximum potential productivity increase would be 50%, not 100%. Factor
in software bloat issues, and it explains why there has been little
contraction in the doubling period at all.

Those expecting contraction when the desktop PC hits human-equivalency
are therefore bound not by the hardware, but the software. If WinXP2025
is as bloaty as all other versions of windows (or linux, Mac,
what-have-you) then there will be no contraction.

And what is the difference between an engineer with a human-equivalent
PC and just two engineers? The only difference is you don't have to pay
the PC a salary, you can make it work three shifts a day, but unless it
is an AI with initiative and authority to do work on its own, it can
only do work during those three shifts that a human has instructed it
to.

This may work with engineers, but if what about management, marketing,
and other functions? If a brilliant 16xhuman PC is sitting on the desk
of an absolute moron middle manager, that PC is still going to be
putting out moron level work, albiet with the appearance of brilliance.

And who says its going to work with engineers? Sometimes getting an
engineer to stop tinkering with a design is neigh impossible, always
adding bells and whistles, gold plating this, calculating that to the
n-th degree. Giving an engineer the ability to do that four times
better isn't going to result in a faster product cycle, it will wind up
with a more delayed cycle and a product that is ever more tinkered
with, and possibly too expensive for the marketing price window.

Mike Lorrey
Vice-Chair, 2nd District, Libertarian Party of NH
Founder, Constitution Park Foundation:
http://constitutionpark.blogspot.com
Personal/political blog: http://intlib.blogspot.com


		
__________________________________ 
Yahoo! Music Unlimited 
Access over 1 million songs. Try it free.
http://music.yahoo.com/unlimited/