Robots are funny, but historically most of that humour has derived from them being the “straight man.” Their ability to be oblivious to the joke often makes them natural comedians. But now Microsoft and The New Yorker are trying to teach a robot to be intentionally funny — specifically so it can help with its popular caption contest, Bloomberg reports.
Since it was introduced in 2005, the caption contest has become a cult favourite among New Yorker readers. The premise is simple: The New Yorker publishes a black-and-white cartoon without a caption and readers send in their best attempts to finish it. The winner’s caption runs in the next issue.
But the contest has become perhaps too popular, and now cartoon editor Bob Mankoff is inundated with 5,000 contest entries every week, according to Bloomberg. And it’s been hard on his assistants. “The process of looking at 5,000 caption entries a week usually destroys their mind in about two years, and then I get a new one,” Mankoff tells Bloomberg.
Who could possibly go through that many entries without becoming completely numb to humour? Microsoft’s answer is: a robot. Microsoft researchers have partnered with The New Yorker to try and build a robot capable of telling which of the captions are funny, and which will elicit only crickets.
Researchers fed The New Yorker cartoon and caption information to the robot, trying to teach it how to tell the difference in humour between similar jokes. Though the “top captions” lists of the human editors didn’t completely align with the robot’s, all the editors’ favourites did appear in the top 55.8% of the robots choices, according to Bloomberg.
Maybe the robot couldn’t pick the absolute funniest caption, but it seems that it could cut out the majority of the awful ones. And even this could save half the workload for Mankoff’s assistants.
Of course, Microsoft’s ambitions aren’t just limited to helping people at The New Yorker have more free time. The researchers told Bloomberg that they hope to one day train the robots to come up with their own jokes, not just know when to laugh. They think this would make digital assistants like Cortana and Siri more “pleasant” to use.