Re: Anti-singularity spam.

From: Kaj Sotala (xuenay@sci.fi)
Date: Mon May 01 2006 - 17:36:53 MDT


>From Eliezer S. Yudkowsky:
> Kaj Sotala wrote:
> > Concrete information about the "ETA" for different transhumanist
> > technologies is something that seems to pop up relatively rarely. I do
> > see a lot of talk about the promised benefits of technology X, and much
> > discussion about why it would be great to have - but often the
> > discussion seems to center more on "why this is great" rather than "when
> > will it be here and why". In part this is understandable, since
> > technological progress is always hard to estimate - but not having *any*
> > concrete estimates for the arrival dates and reasons for new
> > technologies does make all the talk feel an awful lot like just science
> > fiction and hopeful thinking...
> Have you considered that maybe the arrival times of new technologies are
> just *not very predictable*?

Yes - this is why I did say that "in part this is understandable" in my
original post.

My criticism was not meant to encourage people to draw up exact
accurate-to-the-hundredth-of-a-second estimates about when specific
technologies will become available. Instead it was commenting on the
fact that often, any estimates are either drawn up on an entirely
arbitrary basis ("in 20 years" and "in 50 years" seem to be popular,
good-sounding guesses) or just plain non-existent. It may be impossible
to predict the arrival date of a certain technology to a year or even a
decade, but that's hardly a valid excuse for not even *trying* to draw
up rough estimates.

The often quoted bit about computer processing power reaching the level
of the human brain in 2020 or whatever is an example of an estimate
that, while on pretty shaky ground, at least *tries* to project a rough
date based on current data. Better yet might be providing an estimate on
what sort of work still needs to be done on technology X:

* How much has been shown to be possible in theory (the rough idea of
strong AI, with the human brain being a proof of concept)
* How much has been implemented in theory (Drexler's Nanosystems has
extensive analyses about nanotech arguing for its feasibility and
implementations, but so far we don't have the technology to actually
start testing the theory directly)
* If there are theoretical implementations, what needs to be achieved
before we can start testing them
* How much has been actually implemented, either directly or in other
applications that just need some engineering work to modify (there
aren't, AFAIK, any nuclear-powered airplanes in existence, but nuclear
reactors in general have been widely implemented and it wouldn't be a
large step to build such an airplane if we for some bizarre reason
wanted to)
* Where don't we have theoretical implementations, but are confident
that research currently in progress will soon help us come up with some
(and why)
* What are we still entirely clueless about implementing

And so on. This'd probably be better in that it estimates technologies
in terms of things yet to be developed (and may even help directing the
research efforts) instead of attempting to project an arbitrary date
exactly. One could also use *this* analysis to provide (very rough) date
projections - or simply state that there are still too many holes to
make predictions with any reasonable accuracy. In any case, it would be
better than simply saying "no idea" without even trying to do any
analysis, or saying "20 to 50 years" because it happens to sound good,
like often

Foresight's Nanotechnology Roadmap ( http://physorg.com/news4656.html )
would seem, from what I can gather from the press release, like one
example of such an analysis.



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT