Claude Shannon
Claude Elwood Shannon (1916–2001) was an American mathematician and electrical engineer who showed how Boolean algebra describes switching circuits, founded information theory, and moved easily between serious science and inventive play. He is remembered for two works that each changed a field before he turned thirty-three, a 1937 master's thesis that underlies digital circuit design and a 1948 paper that made communication a mathematical science, and for a range of playful machines that anticipated parts of artificial intelligence.
Shannon was born on 30 April 1916 in Petoskey, Michigan, and studied electrical engineering and mathematics at the University of Michigan, graduating in 1936. At the Massachusetts Institute of Technology he worked on Vannevar Bush's differential analyzer, a large mechanical analog computer whose control circuits were built from relays. His master's thesis of 1937, "A Symbolic Analysis of Relay and Switching Circuits," published in the Transactions of the American Institute of Electrical Engineers in 1938, observed that the on and off states of relays obey the algebra George Boole had published in the 1840s and 1850s. A series-connected pair of switches behaves as AND, a parallel pair as OR, and circuit simplification becomes algebraic manipulation, which meant designers could design and verify switching networks by calculation instead of trial and error. The thesis received the Alfred Noble Prize of the American engineering societies in 1939, and it is often called one of the most influential master's theses ever written. He completed a doctorate in mathematics at MIT in 1940 with a dissertation applying algebra to theoretical genetics, a work far less famous than its predecessor.
At Bell Telephone Laboratories, where he worked from the early 1940s, Shannon pursued communication and cryptography alongside his colleagues in wartime projects. He worked on fire-control systems and on secrecy, and a classified 1945 report on the mathematical theory of cryptography was published openly in 1949 as "Communication Theory of Secrecy Systems," which put encryption on a rigorous footing and proved that the one-time pad is perfectly secret. At Bell Labs he also met Alan Turing during Turing's wartime visit in 1943, and they discussed ideas about machines and thinking. His 1948 paper, treated as an event in its own entry, defined entropy, channel capacity and the limits of reliable communication, and gave a generation of engineers a target to approach. Shannon then largely stepped back from the field he had created, and in 1956 wrote a short, cautionary editorial on the danger of overextending the theory to every discipline.
His work on machine intelligence in 1950 combined theory with showmanship. The paper "Programming a Computer for Playing Chess," in the Philosophical Magazine in 1950, set out the basic design still taught: represent positions, assign an evaluation function that scores them, and search the tree of moves, either looking at all moves to a fixed depth or selectively. He estimated the number of possible games at about 10^120, a figure now called the Shannon number, and argued that perfect play by brute force was impossible, so heuristics were essential. In the same period he built Theseus, a magnetic mouse that explored a maze on a board driven by relay circuits beneath it, remembering the route once it had found the goal; it is among the first demonstrations of a machine that learned from experience. In 1956 he joined John McCarthy, Marvin Minsky and Nathaniel Rochester in organizing the summer workshop at Dartmouth where the field of artificial intelligence was named.
Shannon's later reputation rests as much on his character as on his equations. After joining the MIT faculty in 1956 he was known for riding a unicycle through the corridors, building juggling machines, and for working out a mathematical theorem of juggling that relates the time balls spend in the air and in the hand. In the early 1960s, with the mathematician Edward Thorp, he built a wearable computer to predict roulette, an early example of machines used to beat chance. He received the National Medal of Science in 1966 and the Kyoto Prize in 1985, and he spent his last years with Alzheimer's disease. Two misconceptions deserve correction: Shannon did not invent Boolean algebra, and his thesis contributed the application, not the algebra itself; and the label "father of information theory" understates how much of his curiosity ran toward games, machines and puzzles that connect to AI.
Read Claude Shannon as the engineer who found logic inside switches and then found a limit inside noise: two dated insights, 1937 and 1948, that supply the arithmetic of every digital circuit and every communication link, joined to a lifelong habit of treating machines that play, learn and juggle as serious science.