From: Anders Sandberg (asa@nada.kth.se)
Date: Sat Feb 08 2003 - 10:56:30 MST
On Sat, Feb 08, 2003 at 12:15:47PM -0500, Eliezer S. Yudkowsky wrote:
>
> I can fix problems I don't understand. There are no words to convey the
> feeling of omnipotence this generates. The combined knowledge of the
> entire human species is at my beck and call. It doesn't matter whether I
> understand cars. It's enough that someone somewhere does. That knowledge
> is my knowledge for I am of the Children of the Internet.
Very well said.
It brings up an interesting problem: how do we make sure we get more
useful stuff on the net so we can find it? In your example somebody had
written about their car problems - a fairly trivial piece of
information, but he still took the slight effort to write it down.
Similarly for all the wonderfully useful bug reports/"why does this
happen?" for software. But a lot of knowledge is likely still locked
into our heads, not because it is intrinsically procedural and silent
but because nobody bothers to enter it into the web. Maybe the most
important thing about blogging is that people put more of their minds
online (see
http://www.techcentralstation.com/1051/techwrapper.jsp?PID=1051-250&CID=1051-020703Dd9@c
But we still need tools to dig out answers to questions like "Why does
Konqueror crash when I close a window?" and "Are there any easy ways to
generate random numbers with an uniform distribution over a simplex?".
> Has anyone else begun to get the feeling that, wherever you go, you are
> there to represent the Net groupmind? I don't have a wireless wearable so
> it's not really true, yet, but on this occasion the sensation was quite
> strong.
I get it from time to time when I help out my friends and colleauges by
gathering and combining information gathered across the net. We also do
it to some extent in out think tank: by having a couple of well educated
people that can "borrow" ideas and knowledge from each other, we can
write analysis that none of us could do individually (and with Google as
a team member, we also can use the net groupmind).
I guess a sufficiently effective system would start to approach
EarthWeb. Note that this is more of a knowledge pool than real thinking,
and often the information gained is a bit like k-nearest neighbour
algorithms: it suggests what worked for the most similar previous cases.
An intelligent agent then takes this information and integrates with its
own planning abilities, checking whether it is relevant and likely, and
then initiates its own version of the plan. Hopefully reporting the
result.
I'm very much reminded of Sasha's teleological soup paper, where he
describes the collective learning abilities of networked robots or cars.
This is the same phenomenon: we get slightly smarter when anybody learns
(and reports) something.
-- ----------------------------------------------------------------------- Anders Sandberg Towards Ascension! asa@nada.kth.se http://www.nada.kth.se/~asa/ GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y
This archive was generated by hypermail 2.1.5 : Sat Feb 08 2003 - 10:55:24 MST