Magenta, a project from the Google Brain team, released its first computer-generated song, which was composed without any human assistance. While the melody sounds a lot like those old keyboards that came with pre-programmed beats, the first outing is impressive when taking into consideration Magenta is self-taught.
A Google representative told ABC News some drums were added to Magenta's first composition, but the melody is entirely computer-generated.
Whether Magenta's music is any good is another question -- and one Google plans to tackle in the future to help Magenta improve.
"To answer the evaluation question we need to get Magenta tools in the hands of artists and musicians, and Magenta media in front of viewers and listeners," Douglas Eck, a research scientist on the Google Brain team, wrote in a blog post on Wednesday. "As Magenta evolves, we’ll be working on good ways to achieve this."
Machine learning allows artificially intelligent systems to become smarter over time. Google's machines have already shown their prowess -- from translation services and recognizing images to beating the human world champion of "Go," a complex game with trillions of potential moves.
In the case of "Go," Google was able to set up two neural networks. One network would be able to predict the next move while the other would predict the outcome. The program got smarter by learning about games played by human experts while also playing thousands of games between its neural networks, allowing it to better master Go.