Could an AI Chatbot Replace Jimmy Fallon on the Tonight Show?
For the past 5 years, Joe Toplyn has been trying to build a funny robot just like him. It’s a challenge, in part, because he’s a pretty funny guy. The Harvard University alum worked for many years as lead screenwriter for the likes of David Letterman and Jay Leno, collected a small collection of Emmy Awards along the way, and even went on to write and produce for the show. submit Monks. Years later, he wrote a book about what he learned in show business called Writing Comedy For Late Night TVand teaches comedy-writing skills to aspiring comedy writers in New York City.
Then one day, he watched the game show Risk where he watched IBM’s Watson clean house play against human opponents. After studying engineering at Harvard, Toplyn’s curiosity was piqued.
Toplyn told The Daily Beast: “I read some research papers on how Watson works and looked at the schematics of it. He learns that the way Watson works to dominate the quiz show is very similar to the way a joke is formed. It analyzes the prompt, breaks it down into key elements, and finally aggregates what it “knows” about those elements into one answer.
“Well, that’s like picking a subject when writing a joke,” explained Toplyn. “So I developed a feeling that if Watson could do that thingthen maybe there’s a way for machines to do these steps in a joke-writing algorithm.”
All of this happened in 2011. And since then, the world of AI has exploded. In the past few months alone, he says he’s seen a “quantum leap” when it comes to the power and sophistication of AI—especially in large language models (LLMs) like OpenAI’s ChatGPT. After its release in November 2022, ChatGPT laid the groundwork for the kind of hype and hype for a piece of new technology the world hasn’t seen since the release of the iPhone or the launch of Facebook. Even things like Google struggling to play catch up as LLMs began to proliferate and were able to upend entire disciplines such as law, copywriting, and coding.
Now, Toplyn and many others believe that LLM could even lay the groundwork for the development and launch of humorous bots and other AIs capable of creating everything from jokes, funny tweets, even even the entire sketch. It’s a bold goal and extremely elusive given the basic human nature of humor, but it’s not surprising that there have been many attempts at creating a comedy bot over the past few decades. However, if Toplyn and others succeed, they could lead to a kind of disruption that will change the course of professional comedy everywhere — or at least, that’s a potential bottom line for it.
Rise of the Funny Bot
The idea of building a robot capable of humor is not a new one. We can conceptually trace its roots back to sci-fi shows and movies like Lost in Space, interstellar travel, And Which doctor. One of the earliest attempts at an automated joke writing system was Joke Analysis and Production Tool (JAPE) was created by AI researchers Kim Binsted and Graeme Ritchie of the University of Edinburgh in 1994.
The system was hand-coded and relatively rudimentary when compared to the complexity of GPT-4 today, but was still capable of delivering the high-quality word games and puzzles you can find. in a children’s joke book: “What kind of pig would you pass up at a party? A wild borehole”; and “What kind of plants can you wear? A fir coat.”
And so for many years, comedic shows remained mostly within the range of simple wordplay and wordplay like JAPE. In 2013, one separate group of computer scientists from the University of Edinburgh created a bot that can generate humorously similar sentences “I like my X as I like my Y” like “I like my relationships as I like my source: open” or “I like my coffee as I like my war: cold”.
We’ve seen these types of simple prank bots become more popular with the growth of social media—especially with Twitter bots like @dadjokeAPIbotmaking groaning puns like “RIP boiled water. You will be the fog.“or surreal @headlinertroncome up with surreal boundaries-inspired jokes from a fake headliner at a comedy club like this one:
But, of course, comedy is not only pun (or whatever Headlinertron is doing). Jokes can be incredibly complex and complex. JAPE can only scratch the surface of the many layers of social, historical, or even personal context—along with puns and timing—that can turn out to be a great joke. It’s something that everyone has tried to cover up as long as we laugh.
“As a field, the study of humor goes back thousands of years,” said Tony Veale, a researcher in artificial intelligence and creative computing at University College Dublin and author of Your With Is My Command: Build AI with a sense of humour, told The Daily Beast. “Aristotle wrote about humor. A large number of philosophers have put forward a view of humour.”
“But in AI, it’s a matter of understanding and creating text,” he added. “Can we detect when someone is being funny or being humorously insincere? Can I make this machine make things that tickle people’s funny bones?”
But there’s one thing most definitions of a good joke have in common: they must surprise you. “We laughed when we were surprised that an irrationality turned out to be harmless,” explains Toplyn. It’s basically the core of every comedy. Something surprises us, and if it doesn’t hurt us, we laugh.
That’s easy enough for us to wind up with our human mind. After all, people start laughing and understanding humor as soon as they’re born (that’s why babies love peek-a-boo, after all). But for a robot it is much more difficult. There is so much linguistic complexity involved in being funny and creating a sense of surprise and absurdity that the task of creating a bot as funny as a human becomes something serious and difficult. much more difficult.
Or, at least it seemed that way before ChatGPT came along.
ChatGPT to ChatLOL
OpenAI’s hugely popular chatbot has a lot of people asking which jobs will be automated first, which industries will disappear, and how soon artificial intelligence in general will emerge—just to name a few. However, for the likes of Toplyn and Veale, the quest to create a robot capable of creating truly human-like comedies becomes a lot more real.
Even until a year ago, Veale didn’t think bots could create jokes and comedy quite like humans. Now, with the release of LLMs like Bard and GPT-4, he’s not so sure.
“These patterns have become more complex, and we see more emerging behaviors from them,” explains Veale. “These are behaviors that we couldn’t predict and seem outside the scope of a language model.”
On the other hand, Toplyn was extremely optimistic about comedy bots. So much so that he eventually piqued his curiosity about computer humor and took it a step further: He bought his own patent for a comedy bot called Witscript, which he claims is “the first artificial intelligence to create real jokes”.
Using OpenAI’s GPT-3 big language model, the bot is capable of handling any prompt, from news headlines to questions like “Who wins a battle: Mike Tyson or Rock? “; and then roll out jokes that you might tell Stephen Colbert or Jimmy Fallon in their late-night monologues—or at least something close to them.
In a demo for The Daily Beast, Toplyn used the prompt “Donald Trump is under investigation for paying a gag to Stormy Daniels.” Thematic, engaging, and including Donald Trump. It has all the ingredients for something you’ll hear on a late-night circuit today. After a few seconds, Witscript gives four responses:
- What is the big problem? He pays gag for all his mistresses!
- It’s not gag money, it’s just his tip!
- Why did Trump pay a gag to Stormy Daniels? Because he’s Cohen’s man!
- If the payment was a gag, I’m not sure why he hugged her like that.
The bot even chose its favorites out of four. In this case, what’s the big deal? He pays to silence all his mistresses!” Of all the answers, this is probably the closest to being TV-ready. And even if it doesn’t, Toplyn argues, there’s still a lot of value to a funny bot.
“You can think of this as your joke-writing assistant. This is like a comedy writer’s room,” Toplyn said. “The lead screenwriter said, ‘Oh, we need a joke about ‘silent money’ by Stormy Daniels and the writers making the jokes. Some of them will be crummy. Some of them will most of was a joke, in which case some other writer might pick it up and fix the joke.”
Although Witscript currently runs on GPT-3.5, Toplyn says that once GPT-4 is available to him as a developer, he will implement it into the software and hope to get answers. More funny stories from the system.
In her mind, Toplyn believes that a tool like this isn’t just invaluable to, say, Fallon’s writing room. Tonight’s program but it could be a way of democratizing humor and creating jokes in a way we’ve never seen before. After all, not everyone can make good jokes—but you may need them for your scripts, tweets, or presentations at work. A generalized humor bot might be there to help provide humor tools to non-humorous people.
Or, take it a step further: Maybe you want to have your own comedy show on YouTube or TikTok, but you don’t have the money that comes from network television advertising dollars. “What it’s doing is providing comedy to people who can’t afford to hire a human screenwriter or can’t access comedy in any other way,” Toplyn said. “They can’t write jokes or hire a professional writer—but they want jokes.”
He admits that there will likely always be a need for a human in the loop. “I don’t believe in the near future that humans will be completely out of the equation because I think there is still management,” Toplyn says. “This version of Witscript, at least, can make jokes that might pop up on a late-night show, but you need a human eye to say, ‘Yes, that’s what Colbert is going to deliver. and the audience will love it.’”
Of course, as with anything AI-related—especially when it comes to generalized AI that can create award-winning works of art or real-life conversations with users—a robot— Funny boots will ruffle feathers. After all, if a robot is capable of creating something as fundamentally human as comedy and laughter, what does that mean for the rest of us?
When asked if there was anything else Toplyn wanted people to know about comedy, robotics, and making comedy robots, he decided to turn to Witscript for the answer by asking it: “Readers What can The Daily Beast’s creators ask about AI-generated jokes?”
Among its four replies was a very relevant joke, “Why did the AI-generated joke cross the street? To go to the other side of humor!
“A little meditation,” Toplyn noted. “Because, in this piece, that’s exactly what we’re doing: We’re going to the other side of humor.”