There was this wonderful incident of a meeting in England where I first met Seymour Papert, with whom I worked continuously for the next 30 years. And we both had the… had presented pretty much the same mathematical theory of how a... how a certain neural network could learn. And it was reinforcing the connections between synapses in accordance with such and such set of rules. And that alone… that very simple model shows how a machine could learn certain interesting classes of behaviors; not all, but some and... we worked for maybe 10 or 15 years on trying to find to what extent what kinds of phenomena could be learnt by a simulated neural network using this particular kind of learning by reward and… and extinction by not so much punishment, but just by not rewarding it. And we discovered that those particular kinds of neural networks could learn all sorts of interesting and useful things.