Machine-learning can be used for a wide variety of things, but I admit one I never thought of was humor. No, not to tell jokes (though we've seen proof of how humorous that can be through Siri and Cortana), but instead churning through tons of jokes and deciphering which ones are the best, and ultimately, being able to single out unknown jokes as being worthy of being repeated.
Of course, this all ties to a real-world example. For the past decade, The New Yorker has had a wordless cartoon at the back of its magazine, with readers being invited to send in their ideas for captions. Despite receiving thousands of entries each week, assistants at the magazine have been tasked with sorting through them. If this sounds maddening, it is. Bob Mankoff, the cartoon's creator, has admitted that he ends up having to refresh assistants every couple of years because of the mental burnout.
You can probably see where this is going. Working with Microsoft, Mankoff has fed thousands of previous funny captions to the company's researchers to see what kind of accuracy could be had. The results are set to be unveiled on August 13 at the KDD conference in Sydney, Australia.
Given what we've seen from machine-learning in the past, it's not that hard to guess that the findings are going to be impressive. It also tips us off to the fact that machine-learning can be used for many different, sometimes unexpected, things. The sky truly does seem to be the limit.