This year’s winner of the AI Song Contest, in which machine learning was used to create music, was recently announced. “Listen To Your Body Choir” was co-written with artificial intelligence and takes inspiration from the song “Daisy Bell,” the first song to be sung by a computer in 1961. But is a computer program really capable of being creative? “The short answer, right now, is ’no’ or at least ’not yet,’” Chirag Shah, a professor in the Information School at the University of Washington, told Lifewire in an email interview. “While recent developments in deep learning have brought us closer to mimicking human intelligence, we are still far from achieving human creativity.”
Humming Along With AI
For the song contest, teams of musicians, researchers, and data scientists created a four-minute song using AI as part of their songwriting process. “Throughout the song, synthetic elements morph seamlessly into human performance,” the jury wrote in its statement announcing the winner. “Thereby creating an organic synthesis between human and AI, which can be played around the campfire.” Experts disagree on whether AI can create original compositions. One of the many definitions of creativity is the “ability to produce work that is both novel, as in original, unexpected and appropriate, in a way that it is useful,” Teresa Queiroga, a data scientist at the music production company Musiversal, which uses AI, told Lifewire in an email interview. But creating something new doesn’t necessarily mean starting from scratch, Queiroga pointed out. “For instance, when innovatively associating familiar ideas, we are being creative,” she added. “We believe this is where AI systems can be powerful, since they can deal with a large amount of information and combine it in ways that might not be as obvious to humans, resulting in creative findings.”
Smart Doesn’t Mean Inspired
There’s a fundamental difference between intelligence and creativity, Shah said. “While we can often associate intelligence with task performance, we do not have clear measures with creativity,” he added. “Certainly, we are not at a stage near enough to replace human creativity, and we won’t be there anytime soon.” Many artists train by learning to recreate paintings of the masters, Shah said, adding that “in the process, they figure out techniques and discover their own.” Attempts have been made to replicate artistic training for AI systems. For example, an MIT program called “Timecraft” was trained on 200 time-lapse videos of various masterpieces being painted. It then created similar paintings with time-lapsed videos. When shown to people, 90% of the time, they couldn’t tell the difference between painting videos by humans and those done by the program. “So if the ability to recreate a painting by observing and practicing how it is painted is considered creative, this program is creative,” Shah said. “But if we require this program to also feel the frustration, the awe, and the sense of accomplishment that a human artist feels, we are way off.” AI is helping with the technical and mathematical side of music, Roger Firestien, a professor who studies creativity at the Center for Applied Imagination at SUNY Buffalo State, told Lifewire in an email interview. “Bach did the ‘Art of the Fugue,’ and he took it beyond where anyone else can take it,” he said. “The fugue is incredibly mathematical. You can give that mathematical formula to a computer, and it can write fugues like crazy.” Firestien likened the use of AI in music to a human on an electric bike. “You have to pedal, and the motor assists,” he said. “The composer still composes, but AI helps with the additional work like harmony and chord structures. After the thematic material is laid down by the composer, AI can suggest harmonies.” Firestien suggested that no matter how complex AI gets, it may not truly be creative. “Does AI shave or sleep or walk?” he asked. “That’s where inspiration hits. Can AI replace human creativity? Composing is expressing inspiration, and I don’t know if AI can be inspired.”
