Claude Shannon and AI

                                                               Claude Shannon and AI

The reason I am so suspicious of Artificial Intelligence is what I heard from one of my friend who was doing a course on Electronic Communication at IIT Kanpur in 1976-1977. He was taking a course on probability which underpinned what he was being taught in his discipline.Do students of AI learn even the fundamentals of probability? I don’t know. If they do, more power to them. From what I remember, after more than four decades, is that the slower one transmits a message, more is the fidelity that can be achieved, but never less than significant. 

It seems that this did not sit well with Claude Shannon of Bell Laboratories. This too came to me courtesy my friend. Shannon, working on his own time gave the world the bit, the fundamental unit of information that underlies all digital gizmos. I had heard this earlier, but a friend telling me this in a voice resembling authority made me feel small.

In the heading of this blog I did not couple Shannon and AI as a fancy. Shannon showed that “[I]nformation sent from Point A could be received with perfection at Point B, not just often but essentially always.” That “essentially” is the allowance for “arbitrarily small amount of error”; essentially, playing with Limits that we learn in calculus. This too I received at the feet of my friend!

Shannon showed through a demonstration (he worked on it in his house, with the help of his wife) that a mechanical mouse could “learn” from prior experience. If that is not intelligence, what is? If a mechanical mouse is not artificial, then is there anything that can be called artificial? Ergo, any behaviour that exhibits learning, and that too by a mechanical mouse, then artificial intelligence has dawned. This is what Claude E. Shannon did while he was at MIT as an endowment chair professor.

Every college in town, indeed in a village, boasts of courses in AI. Would any student be interested in learning what a humble beginning their field had? Is he or she as intelligent as a mouse scurrying to find morsels of food in hostel garbage cans? Yes, artificial intelligence demands humility as well as intelligence. Neither one is sufficient by itself.

Raghuram Ekambaram