[extropy-chat] "The Singularity Myth"

[Hal Finney]

John Grigg writes:
> > Is everyone here still a true believer when it comes to the Singularity?

Lee Corbin responds:
> In 1990 or so Foresight circulated a questionnaire asking what year members
> believed a nanotech breakthrough (i.e. an "assembler") was most likely.
> I don't remember for sure, but they may have invited speculation about
> a big AI breakthrough as well. Even though I (nor anyone) used the term
> "singularity", (this was still pre-Vinge's use of the term)
> the idea had been in the air for at least a decade. I went on record
> as predicting 2050 for a nanotech breakthrough and 2061 for when (if
> things went well) I'd be unfrozen.  Figures like I was proposing were
> considered very conservative at the time!  So you are right: generally,
> people's dates have receded.

An interesting data point is the FX claim I created in 1996 to
predict the date of awarding of Foresight's Feynman Grand prize,
<http://www.ideosphere.com/fx-bin/Claim?claim=FyGP>.  It describes the
prize thusly: "This prize requires the design and construction of 32
copies each of two nanotech devices: a robot arm capable of .1 nanometer
positional accuracy which fits within a 100 nm cube, and a computational
device, an 8-bit binary adder, which fits within a 50 nm cube."  The prize
itself is linked as <http://www.foresight.org/GrandPrize.1.html>.
(It's amazing that this 1996 URL still works - kudos to Foresight
for their stability!)

This is an extremely ambitious goal.  If you can create 32 of these
devices you can probably create a lot more of them.  0.1 nanometer
positional accuracy will probably require diamondoid; I don't think
there is any way you could get it with floppy materials.  The ability
to make adders and robot arms of diamondoid would imply the ability to
arrange atoms almost at will.  Any technology that can create devices
this intricate, difficult and numerous would probably be what we would
recognize as full-blown nanotech.  The "singularity" would be just around
the corner (assuming that nanotech will in fact lead to a singularity).

Probably mainstream nanotechnologists would say that this goal is
either impossible or decades (if not centuries!) off.  Yet the FX claim
predicts a date of 2022, only 16 years away.  And this date has been
essentially constant in the 10-year life of the claim, plus or minus
three or four years.

Now, FX has had some spectacular predictive failures; perhaps most
notably, 4 GHz CPUs have been predicted to be 6 to 9 months away for
several years now, but it's never happened.  They're still 6 to 9 months
away, according to FX.  I don't know what has gone wrong here, or why FX
has been unable to correct its mistaken prediction.  Clearly, progress
towards faster clock speeds stopped several years ago for some reason,
but FX stubbornly has continued to believe that we will see that 4 GHz
chip sooner rather than later.

Still, such claims are anomalous, and I have seen results indicating that
overall, the predictions have been generally accurate (e.g. about 75%
of the claims trading at 75 came true, etc.).  So at this point we should
certainly take the FX prediction seriously.  OTOH the skepticism of the
expert scientific community has to be given considerable weight as well.
(However I should note that I am guessing about that skepticism, and
it's conceivable that a private IF market run among nanotechnologists
would reveal something different.)

All in all I would say that with the mixed evidence before us, we
should not rule out a nanotech singularity even within the next couple
of decades, but we should also not count on it.

Hal