Computers Aren’t So Smart, After All
I love combing through The Atlantic’s archives. There’s almost no better way of grasping the strangeness of the past than to flip through a general interest magazine from 1960. Here, we find Fred Hapgood grappling with what human intelligence meant in the light of new machines that could do something like thinking. Intelligence was being explored in a new way: by finding out what was duplicable about how our minds work. Hapgood's conclusion was that if you could automate a task, it would lose value to humans. What tremendous luck! Humans value that which only humans can do, he argued, regardless of the difficulty of the task. And that because computers were so good at sequential logic problems, we'd eventually end up only respecting emotional understanding, which remained (and remains) beyond the reach of AI.