AI is taking over Fringe comedy; can robots be funnier than humans?

Artificial intelligence is providing this year’s stand-ups with plenty of topical material. Will it also make them redundant? By Jay Richardson

Can robots be funny? When Alan Turing proposed the “imitation game” in 1950, arguing that a machine would have reached a critical milestone in artificial intelligence if it could fool people into believing it was human in conversation, he specifically mentioned a sense of humour as one of the attributes sceptics cite as distinguishing us.

Turing ultimately expected “the machines to take control”, developing in ways humanity had yet to imagine, but he was sanguine; fellow mathematician Robin Gandy observed that Turing read his landmark paper “with a smile, sometimes with a giggle”.

In the social media era, where anyone is capable of sharing wit online, professional writers of topical comedy are already facing obsolescence, with shows such as Mock The Week and Frankie Boyle's New World Order recently terminated. Now AI is also rising to fulfil the demand for topical quips. In a foreword to Tony Veale's book, Your Wit Is My Command: Building AIs with a Sense of Humor, former New World Order scribe Charlie Skelton admits that the business of coming up with ten gags on a specific subject is by definition formulaic, essentially computational, and that “there's no ridiculousness without logic”.

Courtney Pauroso in her new show Vanessa 3000.

Once the stuff of science fiction, AI is suddenly everywhere, even the Edinburgh Fringe, which needed a global pandemic to transition to online ticketing. Interviews for this feature were transcribed by AI. And even comedy criticism is facing automation.

A year after the “Turing Test”, science fiction film The Day The Earth Stood Still introduced cinema-goers to the alien android Gort. And it's his namesake sharing a stage with Colleen Lavin in her show, Do The Robots Think I'm Funny?, a disembodied robot head, judging and heckling the Chicago stand-up throughout.

Lavin, who also works for a computer hardware company, has programmed her “robot overlord” to analyse the quantity of laughter in her hour, determining whether she should be replaced by a bot.

Much of the heated conversation about AI's impact on the arts is dominated by content theft. Last month it emerged that stand-up Sarah Silverman is suing OpenAI, creator of the popular ChatGPT chatbot, and Facebook owner Meta, for using her memoir The Bedwetter as “training material”.

Darren Walsh acknowledges that, as a one-liner comic, he's especially at risk from the internet's rapacity for “copying and pasting everything, putting it in a book and selling it”. Yet AI is transforming copyright all over again.

Noting that even Sam Altman, OpenAI's CEO, has called it “wildly overhyped in the short term”, Lavin derides ChatGPT as a “plagiarism engine”. “It's almost cannibalising itself because people are putting so much in (of varying quality),” she observes. Garbage in, garbage out.

Comedian Peter Bazely - whose new show is called AI Jesus - agrees. “AI is going to make so many things easier but I think as a comedian I'm safe for now. Chat GPT tends to write in a wedding speech sort of way, a bit sincere for comedy.” He admits, though, that the technology is “more skilled in grammar” and has “maybe” cut “some of the fat” from setups in his show. “But it’s quite happy to say here's an original joke when it's not. It's not always obvious if it's replying with something taken from another comedian.”

Piotr Mirowski is a research scientist in deep learning and AI for Google DeepMind. He argues that AIs with a sense of humour are “almost impossible”, because “the way we evaluate and judge (comedy) is by criteria that's hard for a machine to define and quantify. So much humour is cultural, contextual and depends on minute details.”

Nevertheless, that hasn't stopped Mirowski and 15 of his Improbotics troupe returning to the festival with Artificial Intelligence Improvisation. The 80-strong international group debuted in 2017. But interest is spiking as “many more people have had access and played with AI tools”. Even specialists are struggling to keep up with the technology he acknowledges. “But because we're now being evaluated by huge chunks of society, a lot of mystique has disappeared.”

Trained on the human subtitling of more than 100,000 films on YouTube, mistakes and all, and with thousands of hours of software development, A.L.Ex (Artificial Language Experiment) shares the stage with Improbotics' human improvisers. Interacting through speech recognition and Mirowski curating responses from automated prompts, the robot is the show's star, not because he's a great improviser (although he is improving), but because he constantly surprises his colleagues with non-sequiturs. “Our show is showing a lot of the things humans can do that AI can't,” Mirowski explains. “We are simply using the machine to seed humans with ideas.”

He likens the “strange, crazy” bot to the drunk actor in improv show Shit-Faced Shakespeare, while Improbotics' Sarah Davies describes improvising with A.L.Ex as “adding an extra element of risk, that feeling of running down a hill too quickly”.

For now, the technology is usually the butt of the joke. Yet it still can be incredibly slick, raising new ethical dilemmas. Recording his recent Netflix special, Blocks, Neal Brennan forgot three lines. So he simply used AI to deepfake himself delivering them. Impressionist Peter Serafinowicz has deepfaked himself as Elvis Presley, performing songs written after the King died. Lavin reckons a bot fed the entirety of George Carlin's routines would sound anachronistic, stuck in 2008 when he passed. Bazely, though, wonders “who doesn't want more Sean Lock?” He notes ChatGPT's appetite for absurdity, adding irrelevant dolphins to his joke about having sex on a houseboat, and suggests: “I could take a break while my 'tribute act' builds a bigger audience.”

Darren Walsh says he’s “too scared” to input his 5060 jokes into a private ChatGPT-style engine to see if it can mimic his style. Yet his show 3rd Rock From The Pun uses AI for all audio and visuals, including “stupid pun images” of composite celebrities like Gordon Ramesses, melding the Scottish chef with the Egyptian pharaoh. Saving him invaluable experimentation time, though, the dizzying array of AI options mean that he's still constantly tinkering.

Colleen Lavin, meanwhile, considered using the AI programme Midjourney to design her posters but couldn't justify taking the commission from a fellow artist. Nevertheless, her Fringe brochure headshot is AI-generated, cost just $17, and fooled her mother into thinking it was her.

And here's another thing: A.L.E.x is cute and unthreatening. Humans don't need a massive amount of anthropomorphising to emotionally respond. But Walsh has synced AI-generated images of audiences laughing with artificial guffaws. None of the insanely gleeful but otherwise realistic people in his images have ever lived, a disturbing example of the “uncanny valley”, whereby robots' realism provoke unease or even revulsion whenever they're slightly “off”.

And it's this horror, and the stupidity of artificial intelligence, that Courtney Pauroso is pastiching with her eponymous sexbot character, Vanessa 5000, in her audience-interactive show.

With dishwasher-safe holes, in the uncanny valley where Bouffon clown-meets-robot-meets prostitute, Vanessa is “a slave of the audience, eager-to-please, gonna give you whatever you want. But is she manipulating you?” the LA comic asks. “Is she subservient, a victim or a threat?”

Pauroso has “danced with the devil”, requesting that ChatGPT imagine a conversation between 2001 director Stanley Kubrick and the recently deceased Unabomber, Ted Kacyzynski, and to write a villain's song for the show that was like Ursula's in The Little Mermaid. However, “a tech optimist” friend transcribed a workshopped version of the show into the chatbot, asking it for possible outcomes. Most, she reveals, were “bad. But it gave me two good ideas.”

Ultimately, Pauroso reckons, it’s “so human” to try to sanitise comedy and sex with technology. “And we have so many lonely people. If it makes people happy, maybe this is the answer?”

Colleen Lavin: Do The Robots Think I'm Funny?, Greenside @ Riddles Court, various times, until 26 August. Darren Walsh: 3rd Rock from the Pun, Laughing Horse @Bar 50, 1.45pm, 15-27 August; AI Jesus, Laughing Horse @ the Three Sisters, 5pm, until 27 August. Artificial Intelligence Improvisation, Gilded Balloon Teviot, 6om, until 27 August; Courtney Pauroso: Vanessa 3000, Pleasance Courtyard, 9pm, until 27 August.