The goalposts for what we call “intelligence” keep shifting as AI improves, a respected tech writer argued.
Deni Ellis Béchard wrote in Scientific American that, if you asked someone in 1995 to evaluate modern AI, they might consider it superhuman; large language models today clear milestones that great thinkers, such as Alan Turing, suggested for “real intelligence.” GPT-4.5 passed the famous Turing test this year, convincing interlocutors that it was human, and “the achievement barely made the news.”
Likewise when GPT-4 got a top score on the US bar exam. Stanford computer scientist John McCarthy, an early AI pioneer, made a similar complaint in the 1970s: “As soon as it works, no one calls it AI any more.”