Degrees of Change

Chapter 1A 75-Year Journey and the Diffusion Divide

—Juan M. Lavista Ferres, Microsoft

The concept of artificial intelligence (AI) traces its roots back to 1950, when Alan Turing published his famous paper, “Computing Machinery and Intelligence” (Turing, 1950), in which he asked the now-legendary question, “Can machines think?” and proposed the Turing test as a way to evaluate machine intelligence. That single question laid much of the philosophical and conceptual foundation for AI.

Five years later, in 1955, John McCarthy, who is often called the “father of AI,” coined the term artificial intelligence and convened the Dartmouth Conference of 1956. That meeting is considered the official birth of AI as a field of study (McCarthy et al., 1956).

Predicting the next word through statistical learning also dates back to the 1950s. In the same decade that Turing and McCarthy were questioning whether machines can think and were identifying AI as a distinct field of study, Claude Shannon, the father of information theory, ran famous experiments in which he asked people to predict the next letter or word in English text. He demonstrated that language could be modeled statistically, laying the foundation for later predictive text systems (Shannon, 1951).

In the early days, AI did not learn from data. Instead, systems were built on rules based on expert knowledge. But by 1957, Frank Rosenblatt introduced the perceptron algorithm, the first computational AI model inspired by neurons in ...