There was an event there that was also a high point of my life. I was invited by Bell Labs to give a talk. I found upon arrival that very few people were there, but Pierce who was director of communications and several other people were there, together with their man who was had these models, and they asked me to present my work, which I did. This man at Bell who had had this hierarchy of levels of noisiness spoke up saying that you don't need all this fancy stuff, infinite means, infinite variances. "It's all very bad. I can do it with just hierarchy." And Pierce turned to him, and said with total scorn, "Yes, I know very well that you can represent the motion of the sun and planets around the earth by cycles, epicycles and so on. He represents it by ellipses. He may be right," or "He is probably right," or "It is worth listening to," I forgot the exact words. But Pierce understood my point. Need I say that this made a revolution in communications? Yes and no. Yes, to the extent that the reason why these errors were obtained is that because of not knowing what to do, everybody was transmitting at a very high signal to noise ratio for very few errors. The errors are very few; therefore any error collecting system was on only say one thousandth of the time. If it was on it was overwhelmed, therefore error correction was impossible, therefore it became clear to Bell, to IBM and to everybody that this was very bad engineering, not on the basis of details but of a very bad tree to climb on. You should transmit at a very much lower signal to noise ratio, have many errors, and pay lots to correct them. So, this actually put the wolf away from my door for a long time, because everybody was aware of the fact that by having an understandable, clean, few parameters, short model, I allowed people to abandon a very bad technology in favour of a better one, which I did not develop, of course.