Ludwig van Beethoven’s unfinished symphony is the latest to be given a helping hand by AI. A startup taught an AI Beethoven’s work and his creative process to finish the music, but the move begs the question of where the human work ends, and the computer’s handiwork begins. “I think that the real short- and intermediate-term potential for AI is precisely that it will complement our own creative efforts, not necessarily replace our own creative production as humans,” Kelland Thomas, Kelland Thomas, the dean of the College of Arts and Letters and a professor of music and technology, at Stevens Institute of Technology, told Lifewire in an email interview. “AI tools might be able to recognize what we’re trying to solve and suggest several interesting solutions from which we could choose.”
Musical AI
Beethoven’s unfinished symphony has long frustrated music lovers, but the company Playform AI took on the challenge, using computers to help finish the work. The AI needed to be taught Beethoven’s process for developing various musical forms, including a scherzo, trio, or fugue, wrote Ahmed Elgammal, the director of the Art & AI Lab at Rutgers University, and the team’s leader. “We had to teach the AI how to take a melodic line and harmonize it,” he added. “The AI needed to learn how to bridge two sections of music together. And we realized the AI had to be able to compose a coda, which is a segment that brings a section of a piece of music to its conclusion.” A full recording of Beethoven’s 10th Symphony was released in September as the culmination of Playform’s more than two-year effort.
The AI Revolution
AI technologies have begun complementing human creative work in recent years. For example, photographers use an AI tool in Adobe Photoshop called content-aware fill that allows them to digitally replace parts of a photo by synthesizing entirely new photo-realistic content from scratch. “Performing this same task would have required a highly trained digital artist just a few years ago,” AI expert Matthew Renze told Lifewire in an email interview. “We can also synthetically generate faces, change facial characteristics, re-stylize images, etc.” But when AI produces images or music, is it actually being creative? “That depends on how you define creativity,” Thomas said. “If you consider creativity to be solely a human capacity, then AI can’t be creative.” Human creativity, however, involves a lot more grunt work than we often imagine, Thomas pointed out. The act of inventing usually consists of looking at potential solutions and selecting a next step that both fits the constraints of the problem and that we find interesting. “AI-based creative tools are highly capable of traversing huge search spaces and selecting appropriate next steps which fit given constraints,” Thomas added. “And AI programs can often come up with solutions that seem interesting to humans by using criteria or heuristics which we provide in advance.” The future is likely to bring AI collaboration with humans rather than replacing them, experts said. Thomas points to the recent example of Codex, a new technology that can write computer programs, after being given natural language instructions from a user. “As technologies like this improve, think about something like Codex for film music or for building a video game from natural language,” he added. “This kind of technology can empower human creativity, but it will still depend on us to generate the ideas and vet the output.” Renze said AI is good at performing simple, repetitious, precise, and narrowly defined tasks within a constrained environment. “Humans, however, can flexibly and adaptively solve problems in novel and creative ways,” he added. “This means that we are both uniquely skilled to perform very different sets of tasks. However, when humans and AI collaborate together, they can complete an entirely new set of tasks that neither could do on their own.” Human-AI collaboration in the future will allow teams to diagnose diseases, recommend treatment protocols, and discover new drugs, Renze suggested. “We will also collaboratively design new aircraft, create new building architectures, and invent new products,” he added. “We will also work together to create new music, new paintings, new books, new films, new video games, and other novel works of art.” Correction - October 15, 2021: Keeland Thomas’ title has been corrected from a previous version in paragraph 3.
