From: Ben Goertzel (email@example.com)
Date: Tue Sep 17 2002 - 12:54:33 MDT
> imho, building and connecting 100-400 different idiot-servants (most
> "exponential software") gives the wanted GAI and only 2-4 are needed to
> start earning money. We are starting with chatterbots and qa-systems.
> cu Alex
Well, I don't think so.
I am an advocate of modular AGI design, but I don't think it's enough to
hook together a number of existing effective narrow-AI modules.
Assembling and tuning and tweaking the right narrow-AI modules is hard,
because you find that the narrow-AI algorithms & data structures others have
created are never *quite* right for integration with other narrow-AI
algorithms & data structures....
You wind up creating a lot of your own narrow-AI modules, with a view toward
integration & emergence, inspired by previous narrow-AI work...
Regarding making money, that is a whole different issue. History shows that
profitability is not well correlated with technical innovation anyway --
except in certain particular types of market situation.... The
chatterbot/QA/eCRM space may or may not be one of these special market
situations. Remember Artificial Life Inc., which had very innovative
narrow-AI-based chatterbot tech aimed at the CRM space, and went under about
9 months ago? A cautionary tale, at least...
-- Ben G
This archive was generated by hypermail 2.1.5 : Wed May 22 2013 - 04:00:35 MDT