Patton Oswalt Thinks A.I. Will Get Better at Comedy Faster Than You Think
A lot of comedians believe they have nothing to worry about when it comes to competition from A.I., claiming the best it can write is lukewarm dad jokes and rehashed versions of other people’s material. “In comedy, we have that in human form,” said one comedian. “We call them hacks.” But Patton Oswalt says comics shouldn’t be too hasty in writing off A.I. just yet.
The worry should be about how fast A.I. is learning, Oswalt told Conan O’Brien on the Conan O’Brien Needs A Friend podcast. “When writers make fun of ChatGPT — ‘look how bad this writing is’ — it’s in its infancy right now. It’s growing exponentially. We will get to a point where we won’t be able to distinguish anymore, and that’s going to be terrifying.”
Not that the creators of artificial intelligence are trying to do terrifying things. It’s just what happens with technology. “I forget who said this but any new invention has practical use, accidental harm and then premeditated harm.” Oswalt used an ax as an example. When it was invented, it had a practical use — you could cut wood to build a fire. But you could also drop it and cut off your foot — accidental harm. Or you can murder someone with it — premeditated harm.
“I’m sure that A.I. was looked at as, ‘Wow, what a cool thing! It’s a learning computer,'” Oswalt says. “And now they’re just starting to realize the accidental part of it and then the malevolent part.” What makes A.I. worse than an ax is the ax doesn’t get smarter.
What does this mean for artists with funny podcasts? O’Brien is worried. A.I. “is listening to us talk, and it’s anticipating things and it’s growing,” he says. “Sometimes I think, ‘Am I like three months away from there being a really accurate porno that I’m in?' And I’m only saying this because I did some porno, and I’m trying to cover it up.”
Oswalt finds hope for humanity in, of all places, Spider-Man: Across the Spiderverse. Rather than animation that mirrors reality, Spiderverse “is so over-the-top cartoony that it actually connects even more emotionally, whereas a lot of motion capture that makes it look very human, for some reason you don't link to it.” But because humans are creating through art, he says, “there were scenes where I was almost crying. It’s so emotional. Even though these are very stylized cartoons, they actually connect. So maybe there will always need to be some kind of human element behind it.”
Could A.I. be useful as a writing partner? Oswalt isn’t interested. “I still want the humanity,” he says. “There are scenes in Jaws that don’t quite end right. It’s still a brilliant movie because it’s Steven Spielberg, and it’s a young guy making a movie and you feel it happening.” Could a tool like ChatGPT spit out 20 scene endings to help the next Spielberg have smoother transitions? Sure, but at what cost? “It wouldn’t be the movie that it is.”
O’Brien points out that advanced A.I. could have easily manufactured a virtual shark for the original Jaws, which famously had a mechanical version that didn’t work well. But the malfunctioning fish meant Spielberg had to show the shark much less than he had intended — the very thing that made the film so brilliant.