Smart Paint: How Pindar Van Arman Is Bringing AI to Art With Artonomous
Residing in an art studio deep within Fort Worth, an artist is working tirelessly on portrait after portrait. The artist’s name is Artonomous – and it’s a robot. It’s learning how to paint portraits and its creator, portrait artist Pindar Van Arman, is teaching it. The goal is not just one individual robot painting portraits, but a factory producing art created by AI.
The key is machine learning. The motivation is freeing artists from hours of simple, mundane work on portrait creation. The result could be technology that can paint your portrait and that changes how we understand creativity.
This post is part of a series for AccelerateDFW's Storytelling initiative. For more insights on entrepreneurs in DFW and beyond, visit the Storytelling page.
The story of Artonomous’ creation is also a story of struggle, rejection, and validation. Pindar began in a niche market that hardly anyone was in and that faced a lot of skepticism. A drive to be the best and a belief in the creative potential of AI has convicted him of his project.
AI and Imagination: More Than A Machine
Pindar is currently collaborating on the Artonomous project with photographer Kitty Simpson. You can assist the project yourself by submitting a self-photo here:
Years ago, before Pindar found the artistic success he now has, he wanted to devise a more efficient way to complete portraits. He had a day job, he had children, and it took 8-10 hours to complete a portrait. He wondered if there was a way mechanically he could have the “boring stuff” – paint mixing, laying down backgrounds, proportions – completed while he was away at work. Thus, he could maximize his time on the more creative side of portrait-creation once he returned home.
He made his first robotic helper about 16 years ago. This began a new journey into creating AI and robots that could discern and paint creatively. He wanted to take this helper beyond a basic assistant.
“And I was like, you know what, it can also help me with composition, make sure I got a balanced composition. I was thinking, you know, some paintings sell more than others. Maybe you can learn what is popular with people. And so I started adding small, simple AI algorithms, one after the other, just to like toy with it. And this is like, you know, nothing major, it was over the years, just every couple months I’d try something new.”
As the years continued, Pindar become more convinced that his AI – and any AI, really – was capable of creativity.
Several years ago, Pindar recalls deep learning rising in popularity. Then a few years back a surprising development happened. In the world of professional Go, the abstract strategy board game, the AI “AlphaGo” beat champion Go player Lee Se-dol. Lee Se-dol further claimed that the AI was invincible.
Pindar recalls news headlines running which argued AlphaGo was creative, not just an AI using “brute force.” And the more Pindar looked into machine learning, the more he became convinced that AI could be creative.
“But it's not really creativity, all my AI was like, you know, was algorithms … [that] do specific tasks more efficiently for me. … I've changed my mind I used to think they weren't creative I used to think they were … just procedural or like deterministic. … they always did the same thing over and over again.
“But I looked into deep learning and I quickly became convinced … this is creativity. And I'm not saying that deep learning is an artist, but I realized that deep learning was creative. And creative in many of the ways that humans are creative … and I would start experimenting with deep learning and my robots really took a leap into the world of creativity.”
And that’s the unique pitch here: creativity. The robot and AI that Pindar has developed isn’t just mechanically constructing an image – there is creativity behind it.
This lies in how the robot processes the information for the portrait it is creating. Just like a portrait artist uses creative techniques to recreate your face on canvas, so too does Artonomous. “It has an idea in its mind of what it wants to paint,” said Pindar.
“So this works exactly like a portrait artist would work. It has a face and it’s in its memory, then it looks at the canvas and it decides where to begin. It takes a photo of the canvas and says, ‘Hey, I want to make this canvas look more like this image in my memory. What can I do?’ It looks for the biggest differences. … I call it a difference map or delta map, and then it finds the biggest differences and starts painting away the biggest differences. That's why it always starts with black, because, you know, it's in its memory. It's like, ‘What is the biggest difference between this canvas and this image in my memory?’ The blacks are like, you know, are the biggest difference. It's interesting, if I was to put a black canvas down, it would start on the whites.”
These feedback loops, Pindar says, is the big difference between this project and a printing machine. It operates along creative thought processes – it has neural networks of its own.
The goal is that Artonomous will create evocative, emotional art of its own – that it will move on from copying to creating.
With the creative angle in focus, Pindar began to realize success. As he told me, he was lucky – in the right place at the right time. No one was really into “techno AI arts” and no one was doing much of it, not on the level Pindar was.
He started doing talks, getting on NPR features, gaining support from big corporations, and had a TED talk. He achieved third place and then first in an international contest on making art with robots. He was becoming a public figure on teaching creativity to robots.
Recognition – Is It Even Art?
The road to recognition was not easy, though. He used to get no attention. The issue is that Pindar had trouble getting his AI-driven work categorized as art in the first place.
At a large print-and-paint show in Washington, D.C., jurors refused Pindar’s work. They determined they weren’t paintings, but prints done by a machine. But when Pindar tried to submit them as prints, he was told by those jurors that his works weren’t prints, but paintings.
“I couldn't even compete, the art was so rejected that I couldn't even like be accepted by the art community and at first that bothered me. But then after awhile I was like, wait a minute, this is actually kind of cool, you know, how many other people have other artists saying their art is not art, you know, it's just like, that means you're onto something.”
To hear about how a DARPA Grand Challenge chiseled Pindar's competitive drive, give the accompanying podcast a listen!
For the first five years, Pindar received recognition only in small AI or mechanical art circles. But then when deep learning began to get more popular and mainstream, Pindar’s position become more and more valuable and respected.
“I'm one of the persons. I'm one of the few people that have been doing it for 15 years so I'm sort of like it doesn't matter how bad my early stuff was I was doing it 10 years before any of the new artists.”
Pindar is a pioneer in the field. Whenever a new AI artist joins the community, he connects with them, and they’re familiar with his work. He advanced from being unrecognized, where his art wasn’t even accepted as art, to becoming a figurehead in his field.
AI and Human Value
Pindar’s recognition is not solely positive, though. He gets hate mail for his work. A lot. In short, people are worried about losing their jobs to robots. But Pindar has a different view – an inspiring one.
The tasks that AI would take from us – at least at this stage – are the mundane ones. Enabling AI and automation in the workspace can free us from administrative duties. People would still have their jobs, but they would be able to focus on the creative parts of them, and not have to spend so many hours on the little tasks.
“[I]t could make everyone so much more efficient that we could all of a sudden have two day work weeks. Where it's like, every employee is not just responsible for their own work, they have a bunch of AI assistants and they just have to direct and manage AI assistants. So you show up to work two days, two hours a day, and this is this utopian, Star Trek future that I'm always looking forward to, where we realized that as humans we weren't meant to be in the office doing repetitive tasks 40 hours a week. We realized that as humans we have our passions, we have artistic pursuits, we have each other, and as humans instead of having a five day work week and two days off, we were like, ‘Hey, you know what, we have a five day weekend and work two days and the world is just as efficient’.”
This idea ties back to Pindar’s initial inspiration for Artonomous: making his portrait work more efficient so he could focus his time on the creative, complex steps.
Pindar understands the potential of AI as something that can complement and add value to the human experience. The field of AI is not seeking to displace the human experience, but to add to it.
“Creativity and artificial intelligence, just because a machine can do it, doesn't make our lives any less valuable, doesn't make our creativity any less valuable. A creative AI artist, when one evolves, and when one eventually appears, and it will appear, it's not going to make a human artist any less relevant. It's just going to be another artist. And if the message from that AI artist is interesting that AI artist will do very well. If the message from that AI artist is boring that AI artist is going to be ignored and missed, like irrelevant. … humans are not defined by our work and if AI starts doing work for us, that doesn't make us any less human or any less important.”
Pindar’s work is realizing a new level of human potential. Projects like Artonomous are a testament to the human mind, and an inspiration for what any young, gifted entrepreneur can accomplish with a drive for success.
Humans can not only make art, but also create intelligence that can make art.
Note: The quotations may have been edited for grammatical purposes and to remove chit-chat phrases ("you know", etc.) or repeated words ("and, and", etc.). The audio in the podcast may have also been edited to remove chit-chat phrases, repeated words or long periods of silence.