PETER SCHALDENBRAND, PROJECT LEADER | PHOTOS BY JAMES PAUL

FRIDA is named after the Mexican painter Frida Khalo, though the robotic arm with a paintbrush duct taped to it paints with a slapdashed style more comparable to that of Vincent Van Gogh.

The Carnegie Mellon University team behind FRIDA — Frameworks and Robotics Initiative for Developing Arts — equipped the robot with the ability to create paintings with nothing but a prompt from a user. Given textual, audio or even visual cues, FRIDA uses an artificial intelligence algorithm to figure out how to put the user’s idea down on canvas. Then FRIDA gets started, one brushstroke at a time. 

FRIDA works like a discerning artist. Occasionally, it rinses its brush in a glass of water, dries it on a rag then dips it into a different color before resuming its work. In a few hours, FRIDA, working off only a picture, painted an explosive take of the Smithfield Street Bridge.

The project is led by Peter Schaldenbrand, a doctoral student at CMU, alongside Robotics Institute faculty members Jean Oh and Jim McCann. The team created FRIDA as an academic exploration of the possibilities of AI with the goal to make art more accessible, Schaldenbrand says. 

In addition to images, the user can give FRIDA a text description of what they want to paint, a sound they want to see visualized or any combination of the three. FRIDA takes a photo of the canvas between strokes to assess how to best move forward. Schaldenbrand says by changing course, no two paintings FRIDA produces will ever be the same.

“There’s a billion different definitions of art, and I think it’s constantly changing because as soon as you do define it, somebody’s out there trying to upset that definition,” Schaldenbrand says.

AN EARLY EXAMPLE OF FRIDA’S WORK

The Squirrel Hill robotics lab is filled with ballerina frogs, Albert Einstein dancing, portraits of Frida Khalo herself and starry-night-esque landscapes of Pittsburgh. One painting can take FRIDA anywhere from 1 to 4 hours depending on the complexity.

The initial portraits of a cross-hatched David Lynch were made with a Sharpie in place of the paintbrush, and since then FRIDA has only evolved.

Next, Schaldenbrand says the team plans to equip the robot with hands to bring the capabilities of FRIDA into the third dimension. He imagines the technology being used for the mass production of “hand-crafted” products such as tables and sculptures.

“I think we’re moving into the three dimensions,” Schaldenbrand says. “The methodology that we’ve come up with here, I believe that we can kind of transfer it into sculpting and manufacturing fabrication. I think it’d be really cool to have very unique products created at scale.”

Oh, an associate research professor, says the team is planning to create an at-home version of FRIDA at an affordable price. By making the technology accessible to people with certain disabilities or just beginner painters, Oh says they will be able to express themselves through art they otherwise wouldn’t be able to make.

FRIDA’S PORTRAITS OF DAVID LYNCH

The question that every AI advancement draws is “whose job is at risk?” Text generation programs such as ChatGPT have left writers feeling imperiled, but Schaldenbrand says the goal of FRIDA is to “promote human creativity,” not replace artists.

“I think it’d be fair to call anybody an artist who is trying to do some creative expression,” Schaldenbrand says. “But it’s certainly the person who’s directing the system.”

“I think this is an example of AI working with people properly,” he says. “It’s not trying to replace anybody who is trying to enable new things, and it’s trying to promote human creativity. It’s not an instance where we have an AI that’s trying to do something that’s replicating a person. It’s trying to empower them, and I think that’s the most powerful version of robotics that we have.”

See more of Frida’s work: