February 26, 2024

A comedian and a neural network walk into a bar

Why did the comedian cross the road? To get to the generative AI, of course.

To find out what happens next, watch an amusingly bizarre new video from comedian Ana Fabrega and Cristóbal Valenzuela, co-founder and CEO of AI startup Runway, creator of a tool that generates videos from text, images or video clips. The pair wanted to investigate how artificial intelligence could improve Fabrega’s work.

The resulting footage, which you can watch below, was featured on January 27 at Seven on Seven (7×7), an event sponsored by the New York-based nonprofit Rhizome. The program pairs seven artists and seven technologists to follow a simple, but far from simple guideline: create something new – a work of art, an app, a prototype or whatever – in just a few months. In the video Fabrega, who is best known as the star and co-creator of HBO’s Spanish-language comedy series Los Espookysresponds to AI-generated content created from Valenzuela’s prompts with her own creations.

For example, the Runway manager sent Fabrega an AI-generated song in which she saw the truth in random letters and numbers, and she responded by making a video of her giggling her way through learning the ABC’s lyrics. Valenzuela responded with a strange image of an exaggerated face that looks a lot like comedian Marty Feldman, and Fabrega responded to the oddity with a funny video clip in which she tells the judge in court that she will spend all that free time in prison doing photography. .

Ask, respond, rinse and repeat. The crazy summation of back and forth may not make complete sense, but it shows how AI can lead artists in weird and wonderful directions. And how much potential there is for people and algorithms to work together creatively, with each playing a key role in the outcome.

“Instead of me pushing an AI model, I was asked,” Fabrega said on the 7×7 stage. “Because everything I sent Cris was inspired by whatever he sent me and that was generated.”

This year’s event, the first after a Covid hiatus, took place in the basement of the New Museum in New York and explored how artificial intelligence could change our understanding of everything from love to biology, politics, improvisation and humor.

Before making the video, Fabrega and Valenzuela had never worked together or even met. Rhizome, which offers a platform for digital art, expressly brought them together for 7×7, which started in 2010. Previous participants include Chinese contemporary artist Ai Weiwei; BuzzFeed CEO and co-founder Jonah Peretti; and filmmaker, artist and writer Miranda July.

Fabrega and Valenzuela played with a number of experiments before landing on their human/AI video dialogue, which represents yet another exchange in the ongoing, and sometimes fraught, conversation about the intersection of artificial intelligence and art.

Using audio recordings of Fabrega, Valenzuela trained a voice model to sound like her, but it came across as stilted and could not capture the many vocal intonations typical of her stand-up. He also trained a chatbot to write comedy based on Fabrega’s humor, “learning” it from about 25 pages of her jokes and short stories. Humor is subjective, of course, but the resulting jokes aren’t likely to upset stand-up comedians who worry about AI coming to do their job.

“It just kept trying to take the jokes I had and make them into full sentences,” Fabrega said, “and that didn’t really work.”

See Find the Robot Dog Dance

Other couples presenting on 7×7 included quantum physicist Stephon Alexander of Brown University and comedian and musician Reggie Watts, who together experimented with improvisation to understand how physics, creativity and AI intersect.

David Robert, director of human-robot interaction at Boston Dynamics, collaborated with artist Miriam Simun to explore the connection between humans and non-humans through a performance with Spot, the compact four-legged robot dog. He stomped around the stage next to human dancer Mor Mendel, who was dressed in black and yellow, the same colors as the robot, and mirrored the bot’s frantic movements.

The robot and Mendel circled each other gingerly at times, with the human dancer appearing alternately curious about her automated partner, fearful of it, and attuned to it, a fitting reflection of the complex relationship between artists and AI.

Seeing Spot through the eyes of an artist “is valuable for advancing research and introducing robots into everyday life,” says Hannah Rossi, a field applications specialist at Boston Dynamics who ensured the robot functioned correctly for the presentation. The audience also got to see people through the robot’s eyes, through a projection of what Spot “sees” as it maps its environment, which in this case included the audience and the two-legged dancer.

Generative AI tools like the Midjourney image generator and OpenAI’s online chatbot ChatGPT have artists both excited about the creative possibilities and nervous about the potential implications for their work and livelihoods – and for creativity itself. But taken together, this year’s 7×7 presentations suggest that AI is less a replacement for human intelligence than an alternative set of ways to perceive and synthesize information and the world around us,” said Michael Connor, co- executive director of Rhizome and co-host of the event, said via email.

“While they are clear about the harms and risks that AI poses,” Connor said, “they point to a possible future in which these technologies will be complementary and companions.”

Not funny? Blame the people

That synergy was a key takeaway for Valenzuela and Fabrega. During a Zoom interview ahead of 7×7, they emphasized that art created with AI is only as compelling as the people who help create it.

“Sometimes we hold AI to kind of impossible standards, asking systems to create entire jokes, entire shows, entire movies, or entire songs on their own, and really that’s never been the point,” Valenzuela said. “You can choose which parts you will incorporate into it. It is up to you as an artist how you want to make the most of this.”

Fabrega, who had never tinkered with Runway before, was surprised by how useful the tool turned out to be for aggravating jokes or evoking certain moods.

“It would make really bizarre choices. It would do something unexpected that was really fun,” she said. Other times, she added, the tool “really took away the humor.”

Laughing about the way they labeled AI as an amorphous “it” during their collaboration, the pair concluded that AI itself is not a comedian, at least not intentionally.

If the result of a human-AI team isn’t funny, “it’s not the machine’s fault,” Valenzuela said. “We realized that after we tried to make it funny. It’s not funny. Either we’re funny or we’re not.”

Leave a Reply

Your email address will not be published. Required fields are marked *