The “AI effect” is the recurring pattern by which a task, once a computer can do it, stops counting as real intelligence. Chess, theorem-proving, speech recognition, and image labeling were each treated as hallmarks of intelligence until a machine did them well, after which they were reclassified as mere computation or trickery, and the goalposts moved to whatever remained hard. The pattern keeps the label “artificial intelligence” permanently attached to the unsolved frontier.
The idea is most often credited to Larry Tesler, a computer scientist known for work on graphical user interfaces and for the maxim “Don’t Mode Me In.” On his own consulting website, under “Adages and Coinages,” Tesler addresses the saying directly - and corrects how it is usually told. He notes that his theorem is “commonly quoted” as “Artificial Intelligence is whatever hasn’t been done yet,” but that what he actually said was “Intelligence is whatever machines haven’t done yet.” The difference matters to him: his point was about how people define themselves by a supposedly unique intelligence, so that whatever a machine accomplishes must be reclassified as something other than intelligence. It was an observation about human psychology, not a definition of the AI field.
The phenomenon was named and documented by Pamela McCorduck in her history “Machines Who Think,” and discussed by AI researchers including Marvin Minsky and Douglas Hofstadter. Tesler’s own page is the rare primary source where the person credited with the saying states the exact wording and explains the meaning he intended.
Why business readers should care: the AI effect is a built-in bias in how capability gets judged. A system that quietly automates a task today is often dismissed as “not really AI” tomorrow, which can cause organizations to undercount how much machine capability they already depend on.