Pei, Just a few follow-up comments...
> (2) The above difference also leads us to different research > methodology. To > maximize performance, Ben has been trying to integrate different > techniques, > including reasoning, evolutionary computing, neural network. FYI, there is no more neural-nettish stuff in Novamente. It's been replaced by creative uses of probablistic inference and probabilistic evolutionary learning. > On the > contrary, I have been trying to minimize the design, by using a single > technique to achieve as much as possible. The technique I found is a new > reasoning system, and before reaching its boundary of capacity, I have no > plan to add another technique. I've been trying hard to minimize the Novamente design as well, which is why I've reduced the cognition algorithms to probablistic term logic plus Bayesian optimization, both of which have a common probabilistic foundation. I don't believe the design can be reduced any further (though of course one might find an alternative, simpler design with a completely different mathematical and conceptual basis.) But even with the cognition algorithms reduced to 2, there is still a need for multiple cognitive agents that deploy these algorithms in various special ways, which results in the need for a fairly complex multi-agent software system... -- Ben ------- To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/[EMAIL PROTECTED]
