The evolution of "Artificial Intelligence"

The strangest thing about tracing AI across these sources isn't the familiar hype-winter-hype cycle — it's watching the definition of the target silently migrate. Turkle, writing in the early 1980s, catches the field at its most honest: the hard-core AI position openly aimed at general intelligence, yet daily work amounted to chess programs and robots that couldn't match a two-year-old picking up a block. Bostrom notes something darker — that the pioneers' "speculation muscle" exhausted itself reaching the idea of human-level AI and simply couldn't grasp what came next, let alone the risks. Then Kissinger, Schmidt, and Huttenlocher narrate the pivot that made the whole enterprise viable: researchers stopped trying to encode human insight into rules and instead "delegated the learning process itself to the machines," a move they frame as a philosophical shift from Enlightenment mechanism to statistical approximation of reality. Markoff reveals the original sin embedded in the name itself — McCarthy chose "artificial intelligence" over "machine intelligence" or "cybernetics" partly out of personal pique toward Wiener, then spent decades insisting it was never about human behavior at all, even though the Turing test that underwrites the entire field is literally a test of mimicry. The arc, read end to end, is not a story of machines getting smarter. It is a story of humans repeatedly redefining what "smart" means until the machines' actual capabilities — pattern recognition, statistical inference, graceful degradation — could be called intelligence without blushing.