Hah, being a grad student in machine learning I feel this quite strongly. The popular science perception is that computers are already close to being smarter than us (BS articles about Turing tests don't help...), whereas it seems like if we truly reach computers that "think", it won't look anything like what we're doing now.