by Ken Jennings
Some popular Twitter formats are more or less writing prompts, like the “Damn girl!” pickup lines that were the flavor of the month in early 2013.
demi adejuyigbe @electrolemon
damn girl, are you the wife of a convict serving a long term in a federal penitentiary, because you left before i even finished my sentence
Tony Logan @tnylgn
Damn girl are you a pizza at a Chinese buffet because I’m not feeling it right now but I see you over there doing you and I respect that.
Eireann Dolan @EireannDolan
Daaaamn girl is your name Katrina because my lower 9th just flooded
Yes, I know. The all-time champion “Damn girl” tweet is actually a 9/11 joke of jaw-droppingly poor taste, and is not appearing in this book because I’m not an idiot.III
Other Twitter formats are such rigid templates that they require about as much comic invention as doing Mad Libs. Put the word “Sext:” before a decidedly unsexy clause,IV or add “and chill” to the end of one. End an unpopular or ridiculous idea with “. . . said no one ever!” The most useful formats survive because they provide a satisfying sense of “joke-ness” to what would otherwise be an only mildly amusing observation. They’re like the stock ending of a Restoration comedy, the happy wedding that wraps everything up with a little bow and gets everyone offstage. Take these variants of the old sitcom “pull-back/reveal” joke:
OhNoSheTwitnt @OhNoSheTwitnt
Since Ariel was 16 when she became human do you think she got her period immediately?
Boss: I meant any questions about the presentation.
Jason Miller @longwall26
If it facilitates the consumption of liquids, can it really be called a “crazy straw”?
Lawyer: I think we’ll pass on this juror, your honor
These get a lot funnier when someone figures out a darker way to tweak the format.
Dollars Horton @crushingbort
me: Carly Rae Jepsen’s new album attains an 80s pop authenticity Taylor Swift could only dream of
ISIS captor: hold on its not recording yet
Knock Knock. Who’s There? Joke Fads.
Joke formats predate Twitter by centuries, but that can be hard to remember sometimes because each hot new thing comes and goes so quickly. Who remembers Wellerisms, popularized by Charles Dickens in his first novel The Pickwick Papers? (A Wellerism is a familiar cliché or proverb given some colorful twist by a clever attribution—“ ‘Every little bit helps,’ as the old woman said when she pissed in the sea.”) Who remembers the wind-up doll jokes of the early 1960s (“The Fidel Castro doll: wind it up and watch it turn red!”) or Rich Hall’s invented “Sniglet” words of the early 1980s (“flopcorn: the unpopped kernels at the bottom of the cooker”)? In hindsight they seem like a form of mass hypnosis. How were templates like these ever the hottest thing in jokes?
Folklorists love joke fads because they’re the closest thing we still have to the oral traditions of preindustrial societies. Seinfeld quotes and “You might be a redneck” lists are our epic poems and fireside tales. Honest-to-God academic journal articles have been written about how Helen Keller jokes were a reaction to the mainstreaming of children with disabilities in American schools, and how elephant jokes were a cleverly coded way for whites in the 1960s to worry about the real “elephant in the room”: the civil rights movement. But the real explanation is much simpler and more universal than that: joke fads catch on because they make the mystery of humor seem attainable, for everyone. Each one is a little lesson in how to be funny, or at least “funny.” They’re puzzles. Can you spot the pattern here? If so, you too can crack a joke.
For many years, the vectors of joke-telling didn’t require even that limited creative process. Joke-tellers were mimics, like parrots. If you could repeat the latest naughty yarn about a traveling salesman in 1958, or could say “Yeah, baby!” in an Austin Powers voice in 1998, you could be the good-time office cutup. That kind of thing still exists today, of course, in the form of pass-along social media ephemera: today’s new meme or Photoshop or cat GIF. It’s still easy to make all your friends laugh without actually thinking up something funny. But Twitter is something different: not just an audience to repeat jokes to, but a place where you can use the comic sensibilities of others as training wheels to develop your own.
Funniness starts to seem less and less elusive the more time you spend around funny people. It’s the peek behind the curtain Moshe Kasher and Chelsea Peretti described, like seeing the same magic trick night after night. The first few times you’ll be wowed, but soon you might start to see the strings—especially if you’re shown variations on the same trick. The patterns you learn aren’t strict joke formats; they’re more general comic devices. Reversal is a common one: just take a common situation and turn it exactly on its head. In its most basic form, this is the syntactic flip behind a Yakov Smirnoff–style joke: “In Soviet Russia, TV watches you!” But semantic flips can produce more sophisticated results as well, everything from Rodney Dangerfield one-liners (“I played hide and seek, and they wouldn’t even look for me”) to an elaborate piece of comedy like Dave Chappelle’s “black white supremacist” sketch. Or try applying a situation unexpectedly to itself. Recursion is the transformation behind reflexive jokes as dumb as a “Dyslexics of the World Untie!” T-shirt or as complex as the “Pre-Taped Call-in Show” sketch from Mr. Show that I watched several hundred times in college. Once you see the strings, it’s remarkable how many jokes are constructed on equally simple mechanisms: combining pieces of two disparate situations, for example, or asking “But what if . . . ?” about a common assumption.
Artificial Japes, Stand-up, and Wisecracks
But that doesn’t mean joke production is as easy as memorizing a cheat sheet of five simple rules. If humor really were purely algorithmic, machines could do it, and that’s not happening. My own experience with computer intelligence is limited to Watson, the IBM program that destroyed me at Jeopardy! in 2011, and believe me, there was nothing funny about it. Endless factual recall was trivial for a computer with sixteen terabytes of RAM, enough to store every word in the Library of Congress. But the ability to get a joke—something a four-year-old human can do—was well beyond its abilities. Jeopardy! clues with any kind of conceptual or playful element—anything approaching creativity or humor, in other words—were like kryptonite to Watson.
That’s because humor is one of the hardest problems in artificial intelligence research today. Researchers have written countless joke-telling programs—JAPE, STANDUP, WISCRAIC—but their output is generally as unfunny as their acronymic names.V The jokes they produce are limited to a very narrow domain—mostly punny riddles. (There’s a reason why those were the first kinds of jokes you could master in kindergarten.) WISCRAIC’s best output was a series of homonym puns along the lines of “The performing lumberjack took a bough.” JAPE’s highlight reel includes “What do you call a depressed train? A low-comotive.” These algorithms are also completely clueless as to which of the output they’ve generated actually qualifies as funny, though to be fair that’s true of many carbon-based comedians as well. I’ve tossed off plenty of dumb Twitter jokes that, to my shock, circled the world within seconds of my hitting “Send.” Conversely, I’ve tweeted plenty of things that cracked me up in my head but amused literally no one else. On Twitter, when a joke doesn’t land, there aren’t even glasses clinking and a single dry cough from the back of the room. There’s just the void.
But there’s a very logical reason why computers can’t evaluate their own humor: no one has ever taught them, in a language they can process, what makes jokes work. A Purdue linguistics professor named Victor Raskin has been studying this problem for years, developing a formal language called OST (Ontological Semantic Technology) that will theoretically help computers “understand” the interplay between meaning and ambiguity that makes things funny. His goal is to be able to convert any joke into a labyrinthine “text-meaning repr
esentation,” or TMR—essentially a computer program for humor. OST is so complex that a simple two-sentence quip can break down into a text-meaning representation that’s pages and pages long. Here’s a vastly simplified TMR for the seven-word joke “A man walks into a bar. Ouch!”:
TMR1: walk (agent (person (gender(male)) (age (adult)))) (location (bar))
TMR2: collide (agent (person (gender (male)) (age (adult)))) (theme (pole)) (instrument (leg))
See? See how funny it is now?
My favorite joke algorithm, DEviaNT, was developed at my old school, the University of Washington, in 2011. DEviaNT had it all, from a bad acronymic name—“Double Entendres via Noun Transfer”—to an incredibly limited humor domain. All it was programmed to do was to read a sentence and then determine whether or not it would be funny to follow that sentence with “That’s what she said,” in the manner of Michael Scott, Steve Carell’s character from The Office. But DEviaNT, it turned out, was no Michael Scott. It could only identify 8 percent of the possible “That’s what she said” opportunities, and it notched so many false positives that its jokes were only successful 71 percent of the time.
Just 8 percent! Keep in mind that this program didn’t even have to invent or understand jokes. It had exactly one job: to learn how to use a single canned sentence, the purposely unfunny catchphrase of a TV character. That’s not hard at allVI and it still couldn’t quite nail it.VII
If DEviaNT is the state of the art, we don’t need to worry about funny supercomputers writing our sitcoms anytime soon. Robots will be killing us with lasers long before they start killing with their stand-up.VIII It’s tempting to think that the rise of humor in discourse today is because we all sense this on some level: the ability to crack a joke is one of the most essentially human traits we have left. A good nine-to-five assembly-line job used to be a prestigious career, before machines could do those jobs faster and better than we could. It isn’t just that wages dropped; the whole enterprise started to seem less admirable to us somehow once a robot rendered humans obsolete. Even the knowing of facts, my only marketable skill, seems quaint and silly in the age of Watson and Google. But jokes! The machines aren’t coming for our jokes yet, and so we cling to the small things that can’t be automated away.
Incidentally, the most common rebuttal to the “Chinese room” problem in artificial intelligence circles is “the systems reply.” This argument holds that, even though the person in the room doesn’t speak Chinese, the entire system taken as a whole—the beleaguered occupant plus the program and all the room’s pencils and scratch paper and whatnot—does, effectively, speak Chinese. The room (and, therefore, the computer) knows what it’s doing, even if the individual doesn’t. So computer scientists would say that I am getting funnier—because my system now includes all the mechanisms I learned and joke-sparking connections I made online. In other words, my brain is just one node of Twitter now. Maybe a social media network will be the first artificial brain to achieve joke-telling sentience.
The Big Leagues
Megan Amram @meganamram
At the intermission of musicals there should be a very short football game
Bryan Donaldson @TheNardvark
If Natalie Portman dated Jacques Cousteau they would win celebrity couple nicknaming forever with “Portmanteau.”
If you want proof that Twitter really does work as a comedy laboratory, you don’t have to keep staring into the microscope at individual tweets. At some point, you should be able to see the clinical trials working on your test subjects.
When Megan Amram joined Twitter, she thought it would be a good comedy writing exercise, a way of collecting jokes for the TV pilot script she hoped to write. Many funny tweeters seize first upon the transitory nature of Twitter, and their output is loose and goofy, like texts to a friend. Instead, Amram focused on the medium’s brevity. She took the 140-character limit as a challenge: “Can I think of a bon mot that stands by itself? Can I write a perfect little puzzle?” Her jokes were tiny gems that sparkled in the river of Twitter silt. They felt like discoveries. “There’s literally no way to know how many chameleons are in your house.” “WHY was Mario Kart not called ‘Mario Speedwagon?’ ” Even her “damn girl” joke put on a clinic: “girl are u my neighbor’s wifi? cuz u have a stupid name and im having trouble connecting.”
A friend of a friend saw her tweets, then met up with her at a comedy show. He wanted her help with a TV job he was writing for. On the strength of her Twitter feed, Amram had just been hired for her first gig, and it was the 83rd Academy Awards. About one hundred million people worldwide heard her first televised jokes.
Bryan Donaldson also hopped on Twitter in 2010, to kill time at his IT job for a central Illinois insurance company. Hoping to stay anonymous, he called himself “the Nardvark.” If you were writing his story as fiction, you would never include the fact that he lived in Peoria, the proverbial backwater opposite of the coastal entertainment capitals. Seems a little on-the-nose.
I followed Donaldson’s feed in 2011 and immediately saw that he had a hits-to-misses ratio as high as anyone on Twitter. Tweeting is so easy—too easy, sometimes. As in many sports, it’s all about the shots you don’t take. But Donaldson, like Amram, was irritatingly disciplined and consistent. He only tweeted when he had a great idea. I liked his dry observational tweets, largely of the long-suffering-Midwestern-dad variety (“When my wife gets a little upset, sometimes a simple ‘calm down’ in a soothing voice is all it takes to get her a lot upset” or “Relaxing family vacation. Pick two”) but he also had a knack for political zingers.
It was the topical jokes that brought him to the attention of writer-producer Alex Baze, who in 2013 was hiring writers for Seth Meyers’s new late-night show. He sent Donaldson a Twitter message asking if he wanted to submit a packet. “What’s a packet?” Donaldson wondered.
“I didn’t know the first thing about writing comedy or how it worked,” Donaldson told me. “ ‘Oh geez, this guy thinks I’m a comedy writer and not an IT guy.’ ” But he flew to New York for a fifteen-minute interview, and was offered the job. He didn’t have an agent, a manager, or any comedy experience. His entire résumé was, as he put it, “dick jokes written while sitting on the toilet.”
At the age of forty, Donaldson moved his family from East Peoria to the New York suburbs. The banner on his Twitter page is now a shot of the Manhattan skyline from the roof of Rockefeller Center, the building where he works every day on the Late Night monologue. It’s just like an amped-up day of Twitter: setups will arise from the day’s headlines (now collected for him by a team of writers’ assistants) and Donaldson is on the hook for forty to sixty punch lines. But today, when he nails a joke, he doesn’t just get a few hundred ego-stroking “likes” on his phone. He gets to hear it read by Seth Meyers.
“Twitter is way easier to use to get fired rather than hired,” Donaldson observed to me, but in fact his and Amram’s stories are starting to feel less like one-offs and more like the start of a trend. At Parks and Recreation, Amram worked with Joe Mande and Jen Statsky, both stand-up comedians hired on the strength of their Twitter quips. The original Late Night with Seth Meyers writing staff included Alison Agosti and Michelle Wolf, writers who, like Bryan Donaldson, had been spotted by Alex Baze on Twitter. “Twitter has democratized the process,” Seth Meyers said.
Outsider Art, Without Going Outside
And today Twitter is just one of many ways to become a comedy megastar without even having to leave your bedroom. In 2009 and 2010, UCB improv vets Abbi Jacobson and Ilana Glazer self-produced thirty-three episodes of their web series Broad City, about their own lives as two slacker best friends at play in the Big Apple. Most of the zero-budget episodes were just two or three minutes long, some even shot as video chats. The web series never really went viral, but Amy Poehler was a fan and helped shepherd the show into development at Comedy Central, where it quickly inherited the mantle of Girls as TV’s foremost millennial voice. And B
o Burnham was just a high school junior in Essex County, Massachusetts, when he started uploading his home videos onto the Internet. He would sit in his bedroom with an acoustic guitar or digital piano and sing satirical self-written tunes like “My Whole Family (Thinks I’m Gay).” Within a few years, he was selling out theaters on both sides of the Atlantic and was the youngest stand-up ever to have to his own hour-long Comedy Central special.
In 1982, veteran comedy writer Gene Perret wrote, “There’s a myth around that hidden somewhere in the United States are people much more talented than the big names. . . . The guy in the next office writes better than Neil Simon, but no one has discovered him yet.” But the Internet has now proved Perret’s “myth” to be shockingly close to the truth. Just as the vogue for singing-contest shows brought us legitimate stars like Carrie Underwood and Jennifer Hudson, we now know that the world is full of comedy Susan Boyles, secretly hilarious sleeper agents toiling (or, more likely, smoking weed and playing video games) in obscurity. Maybe Twitter’s best amateur joke writers aren’t Neil Simon, but without any of the advantages of the established LA or New York comedy scene, some of them are already ready for prime time. The ones who make it are the ones who hustle, of course, and it takes time to develop raw talent into real greatness. But it’s a start, and everyone has to start somewhere. Steve Martin used to make balloon animals at Disneyland. Bob Newhart was an accountant for a drywall company.
Even in ancient Athens, a playwright needed a wealthy backer called a choragus to get his comedy produced by paying for costumes, rehearsals, extras, and so forth. So today is the first time in about 2,500 years when it’s possible for funny people to make an end run around the gatekeepers and find a mass audience directly. This is great news for diversity in comedy, since it’s an alternative to the old-boy network that can dominate hiring: which smart-alecky young white guys did they write with at their last job, who did they go to Harvard with. It also empowered everyone—beginners like the Broad City duo and old hands like Louis C.K. alike—to try out the Mystery Science Theater DIY model. They could build an audience on their own and cut deals that kept control of their work far away from network executives, even if that meant taking smaller paychecks. Before its star’s 2017 comeuppance, Louie wasn’t filmed in a suburban Minneapolis industrial campus, but it might as well have been for how little FX, the cable channel where it aired, tampered with its riskier material. This was a series that ended its pilot—its pilot!—with its star telling a three-minute story about the day they put down his childhood dog. Ratings gold!