Can AI understand humor?

Comments · 170 Views

Can AI crack comedy?
Machines are getting better at making us laugh — and challenging our notions of humanity in the process.

When Joe Toplyn, one of the city’s best TV comedy writers, comes into view, the streets surrounding Grand Central Station in New York City are teeming with summertime crowds. The 69-year-old is dressed in baggy shorts, a sun hat, and an obnoxious white T-shirt that reads, “Writers Guild on Strike!”

He continues, “We are demonstrating,” adding that he was recently at Rockefeller Plaza with WGA members picketing the NBC studios.

There is a startling 21st-century plot twist in this story: Toplyn is embracing generative AI for laughs, despite the union’s concerns that studios may employ technologies like artificial intelligence to eliminate jobs.

Toplyn began putting daily news items into a satire algorithm he had developed using natural language processing six years prior. He now shares the results of this Witscript technique on social media. “I sent ChatGPT to joke school,” he claims.

For instance, the @witscript handle responds, “If they keep this up, we might actually get a female president,” to the prompt, “Three different special counsels are currently investigating President Biden, Hunter Biden, and Donald Trump.”

Alternatively, Witscript can respond to the prompt “Kuwait and Lebanon are banning the movie Barbie” by writing, “Lebanon and Kuwait are banning the movie Barbie because they think she’s too Western.” What? In shock, Mattel queries. She is a Chinese product.

There’s a chance that some readers won’t find this funny; humor is a matter of taste. Whether you laugh or not, it’s still vital to pay attention.

When the British scientist Alan Turing created the “imitation game” in the 1950s to determine whether computers could persuade us they could behave like humans, he forewarned that it would be incredibly challenging for robots to pass by exhibiting a sense of humor. The explanation for this is that, unlike chess, humor is a complex example of the confusing and frequently contradictory components of human culture. An Irish computer scientist named Tony Veale claims in his book Your Wit Is My Command that a joke “is like a magic trick” since it ceases to be effective when you logically explain it.

When the British scientist Alan Turing created the “imitation game” in the 1950s to determine whether computers could persuade us they could behave like humans, he forewarned that it would be incredibly challenging for robots to pass by exhibiting a sense of humor. The explanation for this is that, unlike chess, humor is a complex example of the confusing and frequently contradictory components of human culture. An Irish computer scientist named Tony Veale claims in his book Your Wit Is My Command that a joke “is like a magic trick” since it ceases to be effective when you logically explain it.

Consider Improbotics, a theater troupe founded by Piotr Mirowski and Kory Mathewson, two actor-scientists with a Fringe play. On this, a command is given to an AI-based chatbot, which, according to the group’s website, is now built on the GPT-3, GPT-4, and Llama v2 models. The results are then broadcast to human comedians so they can respond.

The humor is based on how people react to these AI outputs and a game where they have to determine whether something is human or not—a modern-day Turing test. “It is pure imitation,” says Mirowski, whose day job is creating AI for the Google-owned DeepMind tech company’s use in things like meteorological forecasting.

But can we actually laugh at this imitation? Or are these automated tools a humorous take on bubblegum, whose artificiality rapidly becomes predictable and unappealing? The issue is far more serious than humorous. The idea of human exceptionalism is now much less secure if machines can enter this inner fortress of culture.

Comments