Reverse Shot A different angle on moving images—past, present, and future

Reading the Room
Chloe Lizotte on AI Homer and the Year in Intelligence

In 2023, artificial intelligence was officially at the public’s fingertips. It was a breakthrough year for advances in generative AI, which was signaled by the November 2022 launch of ChatGPT: a free-to-use AI chatbot that can offer advice on everything from mundane tasks to complex relationship issues. ChatGPT is a generative language model, which means it’s been fed vast amounts of text and trained to recognize patterns within them; the more human-generated material it reads, the more it learns about how we communicate and the better it’s able to read the room as it responds to you.

When you ask ChatGPT to write something related to pop culture, the patterns it recognizes in celebrity speech can be lightly amusing. I asked the bot to share thoughts on the animated film Cars in the voice of Werner Herzog, and it replied with an eight-paragraph essay that began, “In the vast, desolate landscape of animated cinema, one finds the peculiar manifestation known as Cars.” Still, this is a pretty ho-hum mashup. There’s a correlation between the user’s creativity and the end result—that is, the more specific and unexpected you are with your setups, the funnier ChatGPT will be when you put it on the spot.

With that in mind, generative AI enabled a magical, truly out-of-nowhere video series last year—a concert series, if you will. Throughout the year, TikToker @mememusic117 uploaded several videos of Homer Simpson singing popular songs from the ’90s and ’00s. Homer rips into Nirvana’s “Smells Like Teen Spirit” and Limp Bizkit’s “Break Stuff,” has a tender moment with Radiohead’s “No Surprises,” and strains his upper chest voice for Queens of the Stone Age’s “No One Knows.” His family is his backing band—Marge on guitar, Bart on drums, and Lisa on xylophone—and he plays to a grooving crowd comprised of familiar Simpsons faces and characters from blockbuster IP, like Optimus Prime, Superman, Shrek, and the stormtroopers from Star Wars. Notably, they’re all non-human except for the Family Guy characters. The animations are made using Blender, and Homer’s vocals are created with a generative model offered by Voicify.AI. This credit is so prominently featured on all of @mememusic117’s work that TechCrunch writer Amanda Silberberg postulated that this user is posting from the inside to advertise the service.

There is one Homer video that stands apart from the pack: his rendition of Underworld’s “Born Slippy .NUXX,” the ’90s club hit that exploded after its use in the final scene of Trainspotting. Here, Homer is completely alone onstage; the stage lights are dim, and he wears a fedora and sunglasses. If the “No One Knows” clip is funny because Homer’s voice seems mismatched for the material, “Born Slippy .NUXX” sounds like the song he was born to perform, spilling out of him in a seamless legato. Underworld vocalist Karl Hyde brings a stream-of-conscious, talking-singing-chanting vocal approach, which suits AI Homer’s delivery, molded as it is from countless hours of spoken dialogue. “Born Slippy” has a serious edge to it; the lyrics were written to mimic the inner thoughts of an alcoholic in distress, the way they might “see the world in fragments” after a night out, according to Hyde. Thus, lyrical turnarounds like “Drive boy dive boy / Dirty numb angel boy / In the doorway boy / She was a lipstick boy”—in this surreal context, the sounds of human confusion are not so far away from how a glitching machine might speak. Homer’s version, a synthetic soliloquy.

The Homer video is a rare case of AI being used as a tool—a paintbrush—to create something new, rather than as a means to an end, producing an object to gawk at. With some exceptions, AI-generated imagery that catches fire online is not often funny or joyful; it’s usually just worth a chuckle (the Pope in his puffy coat) or plain odd to behold (to keep the Simpsons theme, several artists have rendered how they might look as real people). Then there are questions of copyright and ownership. Last year, an anonymous “songwriter-producer” who goes by Ghostwriter wrote a song, “heart on my sleeve,” for the AI voices of Drake and The Weeknd; its release prompted a debate about using synthetic versions of the performers’ voices without consent or compensation. To be fair, Silberberg’s TechCrunch piece I linked earlier broaches the question of whether Homer’s voice actor might be owed a fee for @mememusic117’s karaoke videos. (In a November issue of the New Yorker, Anna Wiener wrote about the musician Holly Herndon’s initial efforts to create such an artist-centered, economic infrastructure for AI.)

These are important issues that I would be remiss not to raise, but as an art critic, I want to point out that the Drake x The Weeknd song sounds mostly like an algorithmically plausible simulation of a new song from those two artists. It’s not adventurous songwriting, more like sonic wallpaper without a soul. On the other hand, the Homer videos evince no interest on their maker’s part in reflecting the real world back to us, nor in extending a world we already know. They’re more like pop cultural blenders, pulsing familiar references that shouldn’t go together, to do something as pure as throw a party. These concerts are utopian gatherings of IP—fully off the clock, unbeholden to the hellish style templates of bloodless reboots—cutting loose to the club bangers and angst-ridden songs that maybe, just maybe, live within them.

Homer Simpson’s name might hark back to the epic poet—now, in 2023, he is the bard of “.NUXX.” As it was originally deployed in Trainspotting, the song marked Renton’s choice to “choose life,” to break the cycle of addiction in favor of a banal middle-class existence. What does it mean for Homer to do this? As I watch him bare his soul, I feel like I am glimpsing a world I shouldn’t have access to—it is a world that runs without me, seemingly posthuman, but with a uniquely piercing emotional drive. Still, this doesn’t feel as panic-inducing as the idea that AI might replace humans; after all, someone had to come up with this particular collision. Maybe there is another way forward—something less dictated by the rational whims of the human mind, sampling reality to rip a hole in its disappointments.