Ai-Da was built in 2019 and now has an improved robotic arm that allows it to use a regular color palette and brush. Its cameras take a picture of the subject as a reference for painting. But is robot art something humans want? “A hotel corporation that needs to cheaply install thousands of art pieces across its rooms, to add visual flair for its guests, may actually prefer and benefit from AI-generated art if it is easier to procure and costs less,” Rozina Vavetsi, associate professor and department chair of Digital Art and Design at New York Institute of Technology told Lifewire in an email interview. “But an individual may still want to know that the art in their home was developed by a person.”
Robot Paintings
In a recent demonstration in London, Ai-Da became the first robot to paint as human artists do. The robot uses AI to make decisions and create the painting. Each work takes more than five hours, but its inventor claims that no two of the robot works are the same. “In the time of online avatars, AI chatbots, Alexa and Siri, Ai-Da as a robotic artist is acutely relevant,” the team behind Ai-Da writes on their website. “She is not alive, but she is a persona that we relate and respond to.” Ai-Da is only one of many AI programs used to create art. For example, a Paris-based art collective named Obvious uses AI to create art. German artist Maria Klingemann has completed a video installation of Memories of Passersby I, human faces generated by AI, sold at auction. And Google’s AMI program is making waves in the AI art community with a program of art, technology, and machine creativity. There are many ways to create AI art through deep learning, Sneh Vaswani, the CEO of Miko, a robotics company, said in an email interview. He said that generative adversarial networks (GAN) are among the most well-established of these algorithms. “Although GAN isn’t new, its many applications are expanding the bounds of what, and how well, robots can create,” Vaswani said. “And we’re not just talking about paintings and sketches; we’re also seeing GAN applied to music, dance, and other creative areas that were once only thought possible to humans.” AI-created art pieces appear to be created by humans, but are created by computers, usually machine learning or neural networks, Vavetsi said. These networks work by analyzing scores of other artworks, imprinting the artistic styles, elements, and patterns they represent, and generating similar pieces. “By incorporating clever tweaks, the AI can also incorporate creative randomness and elements that might hint at impulse and creativity,” Vavetsi added.
But Is AI Art Creative?
While Ai-Da produces paintings, not everyone agrees on whether those creations are art. “Some might argue AI art will never be fully creative, as it’s just mimicking and spitting out media elements based on fixed technical training,” Vavetsi said. “And that these AI art generators will always require human input to guide them or incorporate flashes of editorial and creative filtering and manipulation to convert it into something truly magical and artistic.” But, said Vavetsi, if AI isn’t already creative, it soon will be. She predicted that AI systems will soon incorporate randomness and noise and take inspiration from many places to “generate the effects of happenstance and the spark of creative impulse.” The US Copyright Office has weighed in on the creativity debate, recently ruling that an AI-generated piece of art cannot be copyrighted because it “lacked the required human authorship.” Dennis Weiss, a professor of philosophy at York College of Pennsylvania, who specializes in the philosophy of technology, contended in an email interview that humans should embrace creative projects such as Ai-Da. “When robots start ‘making’ art, they force us, humans, to think more deeply about what’s involved in the creative process,” he said. “Ai-Da challenges us to think about how human artists have always relied on tools, materials, and techniques to create art.” Update 04/08/2022: The first sentence of this story was revised after publication to better fit the article.
