Photo: Lindsay Eyink
One of the funniest people on Twitter isn’t a person at all. It’s a bot called @Horse_ebooks.
It was originally built as a promotional vehicle for a series of digital books, but in the years since it has developed a life of its own — not to mention a sizable cult following. Its tweets range from the cryptic (“I Will Make Certain You Never Buy Knives Again”) to the bizarre (“No flow of bile to speak of. later. later. later. later. later. later. later. later. later. later. later. later. later. later.”).
OK, it’s no George Carlin, but it’s funnier than most of the Twitter one-liners from your friends and family — and it’s not even trying. It’s randomly grabbing text from e-books and websites.
Hackers Darius Kazemi and Joel McCoy believe there’s a larger point to be made here. For years, people have worked to build machines with a sense of humor — researchers at the University of Edinburgh recently created a program that can actually learn from its past jokes — but Kazemi and McCoy believe these academics are working too hard. Since most people aren’t that funny, the two hackers say, why not replace everyday humor with remarkably simple bots that spew boilerplate phrases over Twitter?
Kazemi and McCoy designed a pair of Twitter bots that do just that. Kazemi’s @amiritebot dishes zingers like “New Music? More like Canoe Music, amirite?” while McCoy’s @ilikelikeilike says stuff along the lines of: “I like my lovers like I like my logic: odd, coherent, not elementary.”
Most of these bot jokes aren’t funny, but that’s the point. These cookie-cutter quips are often called “snowclones.” They follow a specific formula that’s easy to automate and marginally funny, sometimes. In this way, the two hackers are lampooning our incredibly lame efforts to make each other laugh.
“A friend put forth a theory that we should just be making bots for all the bad and lazy jokes people make so that no one has to spend time on them,” McCoy says. “It’s the automated labor theory of humor: let machines do the work so people have the time to think.”
Kazemi, also known for his bot that makes animated GIFs from The Wire, has a similar outlook. “@amiritebot is a commentary on how bad those jokes are,” he says. “I’m interested not in whether computers can be smarter and funnier than people. I’m interested in how people are often dumber and not funnier than computers. I throw these algorithms together in just a couple hours. It’s not rocket science.”
These bots still evolve in ways its designers don’t expect. Kazemi says he tries to minimize the offensiveness of his bots by compiling a list of words that they’re forbidden from using. “If I wouldn’t say it, I don’t want my bots saying it,” he says. But this doesn’t always work. “My algorithms constantly surprise me with their ability to have a dirty mouth,” he says. “I’m constantly adding stuff to [the list]. It’s like: ‘Oh yeah, that’s an obscure slur I forgot about.’”
In a way, this is only appropriate. Trying to be funny, we human beings end up spewing the wrong thing in much the same way.
You may roll your eyes at this little piece of performance art, but the point is well taken. Humor bots are more like us than we care to admit. In 2011, a developer named James Socol created a program called Scottbot could automatically make “that’s what she said” jokes in chat rooms. It would process things typed in the room in an effort to determine whether it could elicit a laugh by responding “that’s what she said.”
Jessamyn Smith saw this bot in action when some of her co-workers installed it in their company’s chat room, and it was about as funny as a live person making “that’s what she said” jokes. In other words, it wasn’t funny at all. It was offensive and repetitive, making between five and ten “jokes” per day.
Smith responded with her own contraption called TalkBackBot. It would monitor the chat rooms, and whenever it noticed a “that’s what she said” joke, it would respond with a famous quote from a famous woman, complete with the counter-tagline “what she really said.” In the end, her coworkers uninstalled their humor bot, and Socol took the bot’s source code offline.
Luckily, the bots created by Kazemi and McCoy are less offensive — and the creators are more, well, self-aware. If you like, you too can get in on the joke: Kazemi has open sourced some of his creations on GitHub and written tips for other bot creators on his blog.
Over the years, Kazemi has built several automated joke machines, ranging from a tech startup pitch generator to RapBot, which generates rhymes in the form of freestyle rap battles. McCoy thinks this style of automated humor could be used to critique everything from news headlines to elevator pitches.
“The general decline in voice and style makes it very easy to make something that at least in passing could plausibility be the thing that it purports to be,” McCoy says. “I think [bots] could be applied to making fun of how lazy so many things are.”
Article source: http://www.wired.com/wiredenterprise/2013/08/humor-bots/