John, > I don't agree, but if you guys want to I'll go along. My problem is I > need to have the comupter "learn" on it's own. My solution is to design > an AI program that can learn using input/ouput. Are you proposing this rabbit to chase? How does this differ from accumulating data? So far, all I've gotten from your descriptions is that you want to amass a huge amount of data and process it in some way. Can we create an experiment to help push this vision forward? What's it look like? David Garner