Twitch’s ‘Seinfeld’ show proves that AI doesn’t have to write comedy

Twitch was banned after viewers flagged the segment for transphobic comments Nothing, forever for two weeks. Hartl, embarrassed and apologetic, said he wanted to be more careful in vetting topics for attackers. But he made no promises that Larry would find anything funnier.

Nanyun Peng, an assistant professor of computer science at UCLA, studies the ability of AI to create “A Unified Framework for Pun Generation with Humor Principles” and “Pun Generation with Surprise”. She says that AI is not funny because it uses a probability model to determine the most expected thoughts, and humor is based on unexpected responses.

That seems like a poor excuse. I’ve worked in a sitcom writers room and writing a lot of comedy is just math. In theory, one should be able to use machine learning to do this. In fact, Peng once taught a model of the rules of comedy that comic writers try to explain. Theories such as the rule of three and the concept of asymmetry. “Our machine was able to generate ‘Greyhound stops to cut rabbit’,” she said. I regretted not laughing, which seemed to insult her. “It’s not that funny. But to some extent. We were delighted to see this result.”

The challenge is huge, says Peng. “I don’t think people really understand jokes. There are no theories where you can use them and then you’ll be a stand-up comedian. Some of it is actually the talent,” she says.

It’s no surprise that comedian Whitney Cummings, who has a robot made like her for a 2019 Netflix special, has an AI telling terrible jokes. “Why are people shocked that robots aren’t funny? Most people are not funny. “The only funny robots are Roombas when they’re stuck under the couch,” she says.

Cummings is generally pro-robot; She even keeps a robot replica of herself in her house. But she didn’t expect him to make her laugh. “Comedy is one of the few things that is most unique to the human being,” she says. “Comedy is about the trauma that comes from human experiences in life and how to deal with it. Robots can’t be hurt.

When I asked Spike Fairston what he wrote Seinfeld Why did he think from 1996 to 1998? Nothing, forever It wasn’t funny, he suggested asking the AI ​​why it wasn’t funny. But when I entered chatgpt it said it was not available because it was at capacity. Interestingly, on the left-hand side of the page, he explained the site’s downfall through an AI chat prompt that said, “Write a comedy diary about ChatGPT status.” The closest thing to humor in it:

Comedian: “I guess I’ll just have to stick to talking to my cat for now. At least he doesn’t have a waiting list.” (laughter and applause)

When I put that to the horse, he replied, “It’s like asking why Spock isn’t funny.”

In fact, when I later managed to log into ChatGPT and asked him why he wasn’t funny, the bot basically said the same thing Fairston did, only with a little more humor: “While AI can recognize patterns and respond based on them, it doesn’t have the humor that humans do.” He doesn’t experience emotion, he doesn’t understand context, he doesn’t understand the nuances of language the way people do.”

However, he was able to summarize this article in a few seconds: I had to throw in a lot of jokes.