reponse to Wlodek
FYI
---------- Forwarded message ----------
From: Ben Goertzel ben@goertzel.org
Date: Thu, Apr 17, 2008 at 12:49 PM
Subject: Re: Quirky question from a wealthy AI patron, Hugo, "China card"
To: Wlodzislaw Duch wduch@is.umk.pl
Cc: Alexei V Samsonovich asamsono@gmu.edu, Hugo deGaris
profhugodegaris@yahoo.com, Moshe Looks moshe@metacog.org, Pei Wang
dr.peiwang@gmail.com, Georges Harik georges@gmail.com, Martin
Pelikan pelikan@cs.umsl.edu, Eric Baum ebaum@fastmail.fm, William
Schuler schuler@cs.umn.edu, dkroy@media.mit.edu, Wlodek Duch
wduch@phys.uni.torun.pl, laird@umich.edu, "Sam S. Adams"
ssadams@us.ibm.com, Selmer Bringsjord brings@rpi.edu,
samsonovich@cox.net
Hi guys,
I agree that a focused project is needed.
However, an obvious problem is that many of us have our own quite
specific ideas about what the focus of the focused project should be
;-)
Thus it seems that, if an AI Manhattan Project were to be funded, one
would want to divide up the work into at least three phases
Phase 1)
Here we just need researchers not software engineers -- although
participation of 1 or 2 folks with software architecture and computer
systems background would be very valuable, to keep planning grounded
in reality.
The goal here would be to arrive at the commonly accepted AI architecture.
At the end of this phase, quite possibly, some of the original
research team might leave the project due to finding the architecture
decided-upon disagrees too profoundly with their own ideas. I
wouldn't consider this ideal but wouldn't be shocked if it happened,
nor consider it an indicator of failure. I think there are many
possible approaches to AGI, and concrete choices need to be made based
on our limited current understanding.
Phase 2)
High-level software systems design
Phase 3)
Begin a long process of integrated, dynamically interactive software
implementation, testing, teaching, theorizing and designing....
I'm not suggesting a "waterfall" software process where things are
designed first and then implemented. Obviously this can't work for
AGI. But I'm suggesting that given the diversity of opinions and
approaches among AGI researchers, an initial phase of coordination and
collaboration to decide on an overall coherent design accepted by a
group of researchers, would be a good idea.
-- Ben
On Thu, Apr 17, 2008 at 9:57 AM, Wlodzislaw Duch wduch@is.umk.pl wrote:
Guys,
I quite agree with Alexei that we need a focused project, and wrote it
already directly to Ben.I'd suggest Web page references to important projects and some info about 5
people you're recommending, perhaps some justification why they would make a
useful team.Wlodek