Neil Gershenfeld's digression

Enter Neil Gershenfeld, stage left.

Neil asks what the structure of the class is. We tell him we're making it up as we go along. Neil says that the language of computation is wedded to classical notions of computing architecture, but its the only language we have right now. It would be nice to write a text book, nice to put together cook book for computational universality. Babbage could end up producing a series of papers like the chaos collective.

Need to learn coding theory (Blahat - information theory and coding theory). Coding theory at least as important as computation (Steve Omohundro has written papers on: do PDEs compute?)

Need to involve Deepa in class.

One of Neil's favorite questions: When will we want to start making machines which signal like neurons? Jerry Lettvin has made a career of studying this kind of thing (Homework problem of contact him and invite him to the group) Neil describes him as an imposing person who has spent lifetime thinking about engineering and biology and signaling. Gil Pratt (Lettvin's mentor) has also done lots of work in neuron-like signaling.

What is the menu of questions that you would ask to come to the conclusion that you should signal like neuron?

(Ben) little knowledge of brain regime. (Neil) lots known about neurons. Neuron like signaling should allow construction of messy computer - low tolerances on components.

Al Bar and Tad Hogg - noisy gradient descent. (question about gradient descent and Q-bits)

Presume that we will write book on subject. Start with this assumption and use it as organizational basis.

Exit Gershenfeld