Rooftop8000 writes:
Yes, but software doesn't need to see or walk around because it lives inside a computer. Why aren't you putting in echo-location or knowing how to flaps wings?
In my opinion, those would be viable things to include in a proto-AGI. They don't lead as directly to conceptualizations which help communicate with human beings but they could be used for conceptual bootstrapping.
It seems that humans (and other animals, to the extent you believe they think) are made by building a general structure capable of learning the nature of the universe, with evolutionarily-discovered optimizations that reduce the learning time. It is not hard to see how "grounding" that process in perception provides a natural sequence of concepts -- spatial regularities leading to spatial relationships and simple mathematics; temporal regularities leading to causality, and so on. From there, analogical reasoning mechanisms have a broad base of material to work from.
If such a learner tries to start with less primitive input -- for example, feeding it text from the web --it's not as clear (to me at least) how it can "grab on" to the primitive elements and build on them. Is there a path from letter sequences to their meaning that can reasonably be learned? I have been following your discussions on this list for some time and finally decided to say something. Although I am not an AGI professional at this time, I like to study the issues in my spare time and think about how smart machines could be built.
----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?list_id=303
