In new research, researchers from Johns Hopkins University show that most people can actually think like computers. The results of the research show that modern computers are not as different from humans as we would like to think. It further demonstrates how the progression in artificial intelligence is continuing to reduce the gap between the visual abilities of machines and humans.
“Most of the time, research in our field is about getting computers to think like people,” said the senior author of the study, Chaz Firestone. “Our project does the opposite — we’re asking whether people can think like computers.”
“These machines seem to be misidentifying objects in ways humans never would,” said Firestone. “But surprisingly, nobody has really tested this. How do we know people can’t see what the computers did?”
In order to test his beliefs, Mr. Firestone and the lead author, Zhou asked people to “think like a machine”. The computers have only a limited vocabulary for naming images so the authors showed people many images that had already confused the computer, to name them from the choices given. It turned out that most people chose exactly as the computer did.
75% of the time, people would choose the same answer as a machine and surprisingly 98% of the humans responded just the way computers tended to.
To up the ante, the researchers in the next step gave the people a choice between the next best guess of the computer and its favorite answer. Once again, 91% of the people agreed with the computer’s answers.
In totality, 1800 topics were tested in the various experiments. The researchers made people chose between 48 choices for what the object was and yet still there was an overwhelming majority of correct answers for it to be a random chance.
“We found if you put a person in the same circumstance as a computer, suddenly the humans tend to agree with the machines,” said Firestone. “This is still a problem for artificial intelligence, but it’s not like the computer is saying something completely unlike what a human would say.”