Re: Steve Harris, "Return of the Krell Machine" (Singularity Article)

From: Dan Clemmensen (
Date: Sat Jan 19 2002 - 12:22:45 MST

Summary: my thesis remains:
     Human/computer collaboration --> "weak" SI --> AI-based "Strong" SI

Ben Goertzel wrote:

> [Dan Clemmensen wrote]
>>This whole argument misses the point: AI may not be the
>>only path to SI. I feel that we'll get SI via human-computer
> I understand the idea.... Francis Heylighen (Principia Cybernetica; Free
> University of Brussels) is a big advocate of this as well, so you're in good
> company.... This is the "global brain" about which I've written a fair bit,
> going back to 1995; and Turchin wrote about it in the early 70's...
> However, my intuition is that there is a *relatively* low upper limit to how
> much human collective intelligence can be accelerated through computer
> technology.

I agree. I do not think that the human/computer collaboration will
continue to operate without an "AI" component. I merely believe that it
may start that way. Let's consider "technical creative intelligence,"
which we define circularly as the collection of qualities that enable
the creation of new technical stuff. If a human/computer collaboration
is more "intelligent" in this regard than today's researchers, then the
collaboration will be able to augment its own technological component
(software) more effectively than today's researchers. So let's speculate
that this relatively low upper bound restricts the collaboration to less
than ten times the effectiveness of an unaugmented researcher. I
speculate that this means that the collaboration can then develop an AI
component in one tenth the time it would otherwise take: 50 years
compressed to five.

> I tend to believe the big breakthrough in this domain has almost "already
> happened". The Net has given us the ability to interact fairly freely with
> people all over the world using our familiar media -- language, pictures,
> and sound. The process will be massively streamlined during the next
> decade, and the Net will turn into an easy place to exchange and collaborate
> on information of these standard types. But will this lead to a qualitative
> increase in collective intelligence? Well, perhaps it will, but a much
> *smaller* qualitative increase than could be obtained if AI's were thrown
> into the mix....

Again I agree, but I don't think you go far enough. You are talking
mostly about existing tools( "language, pictures and sound,") and not
about newer tools (search engines, groupware) that let the computerized
portion of the collaboration do more than just store and transmit data.
Yes, with AI we achieve a second breakthrough, but the first
breakthrough increases our effective intelligence and dramatically
reduces the time we need to reach the second.

> For instance, in biotech, the presence of BLAST and various other Internet
> services and sites has made a huge qualitative difference in how fast (and
> *how*) bioscience can progress. But an AI that could integrate all the data
> from the different sites to make judgments humans can't make (because humans
> can't integrate that much diverse data effectively), would create an even
> bigger qualitative difference...

Well sure, given an AI. That misses the point. If we use un-intelligent
software designed to properly present data, we can let humans do the
integration. This software is easier to produce than AI software. The
DNA sequence search software is stupid and brute-force, but it can
identify possible matches that permit a researcher to see patterns and
make inferences about protein homology. I'm suggesting that we should
try to create a similar qualitative difference in the way we develop
software, and then apply it first to itself and then to AI.

> If there is going to be a *truly astounding* breakthrough here, I suspect it
> will have to somehow involve going beyond "technology as a medium for
> linguistic and visual and acoustic communication." Perhaps, if we could
> directly control some 3D graphical medium via our brain waves, and
> communicate and interact in this way, we could really achieve a hugely
> higher qualitative level of emergent intelligence -- I don't know....
My first little paper on this subject ("Paths to the Singularity," 1996) is

lost, darn it. At the time, I thought that we needed new peripherals,
but I changed my mind shortly thereafter. A good computer display can
present information as fast as a human can accept it, and a
joystick/game controller can accept decisions as fast as a human can
convey them: If you don't believe this, watch a kid playing a video
game. We would all like to believe we think faster than this, but we
don't. In any event I don't think the human/computer interface is an
important constraint on the human/computer SI. It's a software problem,
mostly in the data presentation.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT