Everyone talks about the ability of AI to kill us, but so far it can’t kill on the comedy stage.
On a recent Saturday night, Gene and several other amateur comedians performed short stand-up combos, then delivered four individual segments, inviting the audience to judge human-brained, robot-generated jokes. ChatGPT often featured dad jokes pulled from the internet, making it easy to identify a generic punchline from the original.
“You’ve hit that nerve,” says comedian Gyula Fineman, 31, of ChatGPT who could potentially take people’s jobs, “but laugh it off because the AI is still bad.”
The AI humor tests — which are being replicated in other comedy clubs and researchers — are key to helping better understand the technology, as well as the potential dangers it poses to us. Experts say one of the main dangers of AI is its ability to better mimic and replicate humans, from emotional responses to jokes.
While voice assistants like Siri and Alexa have long punch lines, they’re pre-programmed and non-interactive. ChatGPT and other bots have the potential to take off the internet and eventually come up with their own creative versions.
One such joke appeared in an article published this month by German researchers. Why did the man put his money in the blender? He wanted to pass the time. It was inconsequential, although he did show some creative talent.
But more than 90% of the more than 1,000 jokes she told in the experiment were the same 25, most of them based on puns and puns. Fittingly, researchers have pointed to two common jokes: “Why did the tomatoes turn red? Because he saw the dressing” and “Why don’t scientists trust atoms? Because they make up everything” – were also in rotation at the San Francisco exhibition. .
“ChatGPT hasn’t solved computer humor yet, but it could be a huge leap towards ‘funny’ machines,” University of Darmstadt researchers Sophie Jentzsch and Christian Kersting wrote in their paper.
High-tech companies are seeing their stock prices skyrocket as demand for their products skyrockets. City leaders in San Francisco, where many AI startups are headquartered, are hoping the AI gold rush will revive the local tech scene.
A Washington Post analysis said a snapshot showed 15 million websites had flagged some prominent English-language AI systems. Models like ChatGPT help software engineers generate computer code and can even pass a bar test. But with Hollywood writers on strike at the prospect of technology disrupting their careers, the demos and research suggest the tech could take some time to catch up.
Naomi Vetter, assistant professor of robotics at Oregon State University, studies how robots can help humans in health settings, such as guiding people through physical therapy exercises. Beginning in 2018, Fitter wrote comic book routines for a robot she named John and sent him on tour in Los Angeles.
Jon the Robot uses artificial intelligence to decide where he should jump next in his human-like script. Vetter says John can say the joke failed and then laugh at the joke failing, trying to fix the interaction. “Maybe it’s the audience laughing at them, trying to figure out why they don’t like the joke,” Vetter says. Most of the time, when the bot tried to save the joke, it improved audience reaction, a result Vetter found “promising.”
“You were a huge fan,” John says. audience In a 2020 YouTube video. “If you love me, book me and help me get back to work. »
Humor usually requires a subtle mix of the mundane and the absurd – and so far ChatGPT lacks the brevity and creativity to be funny, say comedy experts. Except, of course, when you’re “hallucinating” or volunteers provide inaccurate information.
“The humor comes from the poor quality of the AI,” says Victor Treviño, a 33-year-old engineer and comedian who splits his time between San Francisco and Los Angeles. Treviño produces a show in which comedians perform standing combos while performing using an AI image generator on stage.
“It’s funny what an AI image generator would assume of a person,” he says, “like putting them in a 1960s scene or giving them extra hands. “It’s like a playground for me. »
When David Isaacs, director of screen and television writing at USC’s School of Cinematic Arts, asked ChatGPT to write a movie scene where a man struggles to tell a woman he like it, the program launched three pages with little to no flair. . He noted that this could be a way out of writer’s block.
However, he added: “It gets me somewhere. “He took me out of the tyranny of the open page. »
Some comedy writers argue that artificial intelligence is needed to perfect the art of levitation. Years ago, while reading about the loneliness epidemic, former late-night sitcom and TV writer Joe Toblin discovered that people might eventually become more receptive to make-up companions — and those companions would need help. ‘a sense of humor.
Toplyn, who has degrees in engineering and applied physics, used artificial intelligence tools to create a joke chatbot he called Witscript. Imagine a more concise and slightly silly version of ChatGPT. Like Isaac, Toplyn sees the potential in misfires.
“It might give you an idea for another joke if Witscript turns out to be a joke that’s not quite there,” he says.
Artificial intelligence is a frequent topic of conversation in the Bay Area — so it’s only natural that it could make its way onto the comedy scene.
Stroy Moed, a 35-year-old comedian, came up with the idea for an AI-themed comedy night called LaughGPT after hearing audience members from another show talk about high tech. He said LaughGPT sold out faster and with less effort than Moyd usually put into marketing.
“It was just an experiment,” Moyed said before the night’s final back-to-back performances, which drew tech enthusiasts in their 20s to late 60s.
When it comes time to test the ChatGPT lines in San Francisco, comedian Feynman turns into a generic joke voice – a bit nasal and wobbly – that makes the joke sound unoriginal.
“What’s wrong with airplane food?” “Said Feynman. “The flavors are very simple. And the prices are too high. »
Towards the end of the show, the audience is finally baffled.
“My girlfriend broke up with me for making too many Linkin Park references,” says comedian Joseph Anolin, 42, as he shoots his set. “But in the end, it doesn’t matter. »
” You! The audience screams.
“It was ChatGPT, baby! »
The crowd is screaming louder than all night.