PERSON
Arthur Samuel
The IBM engineer who coined the term “machine learning” in 1959, taught the first computer to surpass its own programmer at checkers, and posed the deepest question of the AI age in its cleanest form: a machine can learn without understanding, which means understanding is something more than learning.
Arthur Samuel is the founder most people in artificial intelligence have forgotten. Born in Emporia, Kansas, in 1901 and trained as an electrical engineer at MIT, he arrived at IBM in Poughkeepsie nearly fifty years old and spent the next decade doing something no one had done before: teaching a machine to improve itself by playing checkers. The two words he put into print in his 1959
IBM Journal of Research and Development paper—“
machine learning”—now sit at the center of the most powerful technology of the century. Samuel did not theorize about learning from a distance; he built a program that stored what it had seen, generalized from experience,
played itself ten thousand times, and came to play a better game than the man who wrote it. The questions we now ask about superhuman AI, about
memorizing versus generalizing, about
reward and self-improvement,