From: Brian Atkins (firstname.lastname@example.org)
Date: Sat Mar 18 2006 - 14:21:39 MST
Philip Goetz wrote:
> I would have more sympathy with you if you pointed out something in
> the article that was actually wrong, or even debatable.
I don't intend to get deep into this, but here's some quick comments:
On Scientific Rigor
Section 2 here is mostly just sour grapes... I mean really, complaining that
this book had the audacity to use data that you yourself compiled. But as an
overall point that the book indeed is not a scientific treatise, well duh. Of
course it is not. It's not intended to be that kind of document. I'd like to see
a whole area of science devoted to studying this stuff, in fact SIAI is kicking
around the idea of splitting off an effort into helping launch such an effort,
but this book is not intended to fulfill that need. It's more of a wake up call.
Section 3, complaining about where Kurzweil picks for the knee of his graphs,
and that he is "arbitrarily" choosing the Y axis numbers in a biased way, just
seems silly to me. Kurzweil tends to choose orders of magnitude for the Y axis
of his graphs in order to more effectively show the types of changes we find
ourself heading into, but these complaints about his presentation choices don't
change the underlying facts at all... the graph is heading upwards, and we
happen to be alive at a moment where the particular Y axis points we expect to
hit soon will have far ranging effects in real life.
Where Are We on the Curve?
A big problem here where he brings up Moore's Law and claims Kurzweil agrees
"Moore’s law will reach the end of its S‑curve before 2020" and comes up with an
expectation of us being a factor of 30 away from the ultimate limit of computing
This seems to be a fundamental misunderstanding of one of Kurzweil's points that
computing power increases through multiple paradigms, and that he fully expects
there to be additional paradigms past the current one of silicon chips. In fact
Kurzweil spends some time later in the book, which apparently this guy didn't
get to, where he does attempt to point out what the maximum natural limits of
computational power might be, and they are way way beyond 30 times our current
The other graphs have little relevance, other than showing yes indeed single
paradigms eventually top out. Again, missing the main point.
Moving on to the "On The Singularity" section
I have no idea where he is pulling 6x10^23 ops on page 71 of the book from for
expected supercomputer speeds in 2045. There is a graph on page 70 of my book
that shows something in that area FOR $1000, but that is not a supercomputer.
Additionally we have this guy again then attempting to impose his
single-paradigm S curve limits. Kurzweil does back up later in the book his
guesstimates of computing power much higher than the 10^25 range, but this guy
either didn't read it or doesn't want to respond to it.
Section 2: looks like more imposition of this guy's single-paradigm viewpoint.
All four summary bullet points seem to be utterly destroyed by the
misunderstanding of shifting paradigms, and S curve upon new S curve, and the
fact that in many cases the curves Kurzweil shows are meta curves that subsume
these multiple paradigms into one graph.
-- Brian Atkins Singularity Institute for Artificial Intelligence http://www.intelligence.org/
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT