Google’s AI Comes Up With “Never Heard Before” Sounds

Jesse Engel was playing an instrument that was more or less between a Hammond organ and a clavichord—20th-century rhythm and blues crossed with 18th-century classical. As he dragged the marker on his laptop screen transversely and further back and forth, the instrument resulted to be somewhere between a clavichord and a Hammond—but previously it was 15% and now it is about 75%.

According to Cinjon Resnick—one of the Engel’s equals—this wasn’t like playing the two instruments simultaneously. The machine and the software were delivering something unusual layering the clavichord’s sounds on the top of those Hammond’s sounds. Completely news sounds were produced utilizing the notes’ mathematical characteristics emerging from the 2 instruments. And this miracle can be experimented with thousands of different instruments, from violins to balafons, thus creating myriad new sounds apart from already existing ones—all because of artificial intelligence.

Resnick and Engel are a part of Google Magenta. Google Magenta is a small squad of AI researchers within the Internet major building computer systems, which are capable of making their own art. This is also their most recent project and is known as NSynth. Further, the team will widely reveal the technology this week at Moogfest—the annual music, art, and technology festival—that will held in Durham, North Carolina this year.

NSynth was first disclosed by Google last month in a blog post. The main motive is that NSynth will be providing musicians with an exclusively fresh range of tools for music making.

Being a part of Google Brain—the company’s central AI lab—Magenta is where a small group of researchers is looking into the boundaries of neural networks and other structures of machine learning. The complicated mathematical systems that are capable of learning tasks by evaluating huge data are known as neural networks. In recent years, these networks have proved themselves to be a extremely effective way to recognize faces and objects in photos, identify commands given to smartphones, and translation from one language to other. And now the Magenta team is taking efforts to implement this idea for utilizing neural networks as a medium to teach machines to create new music forms and other art.