George Miller and Joe Licklider gave me this laboratory in the basement of Memorial Hall. And then, it must have been in 1949, I was working on this theory of making a reinforcement based... a Skinner based reinforcement learning theory, except it would have a random network of neurons and a way of changing the simulated synapses with little motors and potentiometers and things like that. So I designed this machine, which would be a randomly wired neural network learning machine. And then I... graduated, did a little thesis in mathematics about fixed point theorems on spheres that I mentioned before, and went to Princeton, but in the meantime I wrote a little proposal to George Miller about the theory of this neural network learning machine and maybe we should build one and see how it worked. And then... it must have been the summer of 1952. When... one summer was at Bell Labs and one was... [sic] so this must have been the summer of 1951 and George Miller got some money to build this machine from the Air Force Office of Scientific Research, which was a laboratory somewhere west of Boston. And he had convinced some colonel or bureaucrat or something to get a few thousand dollars, which was an immense amount of money in those days. And... I built... oh and I... and I had a friend, another graduate student at Princeton in physics named Dean Edmond... Edmonds, who was good at electronics. Both of us were, but he was... he knew even more about electronics than I did. And he offered to help build this thing.