X-Message-Number: 5366
From:  (Robin Hanson)
Date: Wed, 6 Dec 95 17:31:50 PST
Subject: Which First, How Fast, and How Sudden?

I just caught up on a weeks worth of cryonet posts and I'd like to say
that I'm impressed with the high quality of recent posts.

Though many posters write as if disagreements were large, I feel as if
I largely agree with most everyone who's posted.  I must be straddling
that young-buck vs. old-fart fence.

It seems that everyone pretty much agrees that *eventually* we will
get V.R., nanotech, revival, uploads, and even A.I.  What folks
disagree about is which will come first, how fast, and how suddenly.
And most folks acknowledge that there is a lot of uncertainty here.
On these subjects, here's what I think.

ON WHAT'S FIRST: As an AI researcher for nine years, I think that
problem is extremely hard, harder than nanotech or uploading, hard
even given nanotech and uploads.  And I'm with Platt in seeing
meat-revival as harder than uploading (I expect to be revived directly
to upload form).  Space colonization also isn't anywhere near
economical soon.  And virtual reality will grow mainly to the extent
it helps folks do "real" stuff - the entertainment and navel-gazing
market is limited.

ON HOW FAST: I think historical trends give us a good baseline for
extrapolating average rates of progress.  It is easy to overestimate
future growth rates if one just looks at how long it should take the
best people, if they were well-funded and managed, to develop a
widget.  One shouldn't ignores all the slower processes needed to
integrate that widget into the rest of the economy.

We've managed to "surf" a wave of accelerating progress for several
centuries without too many spills.  Our mechanisms for projecting and
adapting to progress have accelerated along roughly fast enough to
keep up.  The horizon of our foresight is nearer now in
wall-clock-time, but the more important clock is set by the social
processes by which we adapt to change.  If our reaction times increase
enough as the fog before us thickens, and if our headlights get better
fast enough, then we needn't brake our speed for fear of needing time
to swerve to avoid an upcoming problem.

ON HOW SUDDEN: There have been semi-sudden transitions in history,
such as developing the A-bomb.  Our institutions of adaptation turned
out to be enough to handle this (so far), but it is easy to imagine
that they might not have.  I think analogies to this sort of fear are
the heart of visions of "singularity": changes so big and fast that
things get "out of control".  But while these big fast changes do
happen, they are rare, and so I don't think we should focus our
planning around such scenarios without good reasons for believing in
some specific exception.

And this is where I think the singularity advocates usually drop the
ball.  It is not enough to point to how far change will eventually go,
or to processes which support each other in accelerating change.
Folks must instead lay out more specifically why they expect sudden
change at some special point.  I don't think nanotech advocates have
met this standard in defending Drexler's sudden transition scenario --
vague references toward "design ahead" aren't enough.  And Vinge has
done even worse in explaining his AI-driven singularity.  I (of
course) think I've done better in my arguments for a relatively sudden
uploads transition in my "If Uploads Come First" article.

Robin Hanson    http://www.hss.caltech.edu/~hanson/


Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=5366