On 9/7/06, Fredrik Heintz <[EMAIL PROTECTED]> wrote:
>
> My post was only a mere scratch on the surface... It's a huge topic,
> so I don't expect you to solve anytime soon :)
>
> My post was only a mere scratch on the surface... It's a huge topic,
> so I don't expect you to solve anytime soon :)
Well, I'm not even trying to solve everything by myself. My philosophy is to provide a kind of superglue to plug the modules together.
There are a huge number of researchers doing the following stuff:
-- vision / perception
-- acting / planning
-- natural language
-- inference with uncertainty
-- pattern recognition
-- machine learning
etc etc
What if we can provide a common substrate that they can work on? I think the crux of the matter is knowledge representation and the cognitive architecture.
For example, in vision (which I have spent some time on), there is a lot of overlapping with pattern recognition and probabilistic inference. So I'm trying to convince some vision researchers to re-formulate their work in a standard logical form. They may lose some flexibility, or have to use some suboptimal methods, but what they gain is the participation in an AGI project and relief of some burdens that are not central to their research.
In your planning field I guess the situation is similar. Planning requires inference, pattern recognition, declarative knowledge, etc. Part of the reason why you think planning is so complicated is because you're also dealing with a host of cognitive problems under various headings.
YKY
To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
