Beat by Beat: How Sony is Using AI to Replace Drummers

The machine learning algorithm was trained on hundreds of tracks.

Sony revealed this week that it has created a machine learning drum machine — it seems that even musicians might have to look over their shoulders at AI, with the wealth of music-generating solutions being created.

Sony's new drum track generation algorithm is capable of creating kick drum parts that fit contextually into instrument tracks.

RELATED: 3 START-UPS LEADING THE MUSICAL AI REVOLUTION

"Musically plausible" drum patterns

As Engadget reports, Sony's artificial intelligence is capable of building "musically plausible" kick-drum tracking patterns based on existing instruments used in songs.

Advertisement

Sony's researchers trained the machine learning algorithm on 665 different songs from genres including rock, pop, and electronica. The songs all featured rhythm sections but were missing kick-drum parts.

Advertisement

 The algorithm creates drum patterns based on the instrumentation, song tempo, and speed changes throughout the song it's trained on.

Advertisement

Various samples of the AI-generated drum tracks can be found in Sony's press release detailing their work.

Advertisement
Beat by Beat: How Sony is Using AI to Replace Drummers
Source: Sony

Other AI musicians

While Sony isn't about to write off all of its signed human artists and replace them with AI alternatives, this research does point to the remarkable effectiveness of AI-generated music.

Advertisement

Google's AI project Magenta explores "the role of machine learning as a tool in the creative process" and generates human-like music. Their tool is open source and can be used by anyone.

Advertisement

Facebook has used AI to convincingly convert a whistle tune into an orchestral overture

Advertisement

Startups, meanwhile, like Amper have created AI tools for generating music. Their tool was created so that companies that need music could create and personalize it in a really short time rather than scrolling through countless stock tracks in order to find what they need.

Other companies and programmers have created machine learning solutions for generating music with algorithms. The solutions don't quite compare with the human touch — for the time being at least.

Advertisement