Can a Robot Write a Funny Caption?

Editor’s Note: This article previously appeared in a different format as part of The Atlantic’s Notes section, retired in 2021.
Sometimes they’re amusing. Often, they’re mystifying. Occasionally, they’re transcendent. Whatever else they are, though, cartoons remain beloved components of a fellow magazine of ideas that shall remain nameless (kidding—it’s the New Yorker). Also beloved is the caption-a-cartoon contest the New Yorker opened to readers in 2005. The contest is, for varying reasons, extremely difficult to win. (The great Roger Ebert famously won it, but not—according to the tabulations of Bob Mankoff, the New Yorker’s cartoon editor—until he’d entered it 107 unsuccessful times.)
Because of that, figuring out how to game the caption contest has become something of a parlor game. (An exceptionally nerdy parlor game, but hey.) But the latest attempt to game the system has less to do with cheating than with automating. The tech culture site The Verge has built a robot that, it claims, might—might—help you to wry-caption-y glory:

We’ve collected all the first, second, and third place winning entries going back to when the magazine introduced the competition in 2005—all 1,425 of them. Then, we ran them through a Markov text generator program that analyzes the winning captions and generates new, randomized entries that echo the original set.

Click on “recaption” in The Verge’s generator, and a new caption will emerge. You can then, should you choose, submit that caption to the New Yorker. And then your fate will be up to Mankoff and the other powers that be at the magazine.
Which: gah. Where is the artistry? Where is the creativity? Where is the wit—that bone-dry, desert-arid wit? (Also, perhaps: Where are the New Yorker’s lawyers and their cease-and-desist letter?)
The good news, though, for the non-robots among us: The captions The Verge’s bot generates—at least the ones it served up to me—are … non-sequitors rendered in a language only approximates English. “You’re cut out for a night,” one of them reads, under an image of a man looking like he's been desert-island-stranded crawling into a bar. “Dibs on the furniture,” goes another, for the same image. “I’m going with the oppressor,” goes another.
You can take that nonsense, certainly, as an intricately wry commentary on the actual cartoons and captions the New Yorker runs. You could also take it as a sign that the production of droll cultural commentary is one of the few jobs that robots can’t, as of now, take from us humans. Wit, fortunately, hasn’t (yet) been automated.