Dadabots

Images: Bot portraits via Github
Issue: #2, Summer-Fall 2015.

Algorithms coded and deployed by Zack Zukowski and CJ Carr explore the possibilities of generative music composition and remix.The bots scour music online, post remixes to their own profiles and leave a trace by commenting on the original source. In this correspondence, Zukowski outlines the project and discusses the potential benefits and threats of a bot culture in music.

Dadabots

What are the origins on the Dadabots project— was it an extension of your existing research?

Dadabots began during “Music Hackday Boston” at MIT in 2012. This was a 24-hour computer programming event hosted by Echonest, a company specializing in using machine learning to create music intelligence databases. CJ and I began to fantasize about designing “artificial remix artists” which where fully autonomous using the Echonest API. Soundcloud was a perfect place for these artificial artists to live, because our bots could have the capability of socially interacting with human artists around the world while finding source material to remix without any supervision from us. The code can run automatically off of a computer or web server and our first wave of Dadabots are all live on Soundcloud taking other artists recordings as the source for their creative output.

 

Our idea does have roots in the work of others. David Cope, a music scientist from the University of California at Santa Cruz, conducted experiments in music intelligence by training computers to imitate the styles of classical composers like Mozart using Markov Chain algorithms for pattern learning. Cope’s algorithms became so successful at writing in the style of other composers that he was supposedly able to convince some music scholars that he discovered a lost Mozart piece.

 

With the ever-increasing amount of music being published exclusively online, we noticed it would be very easy to remix music by automatically finding suitable content on social media sites and manipulating it. We decided to take David Cope’s concept a little further and build bots capable of re-composing other artists’ songs, generating the visual artwork and commenting on music from the original artists page to inform them that they have been remixed. The goal was to take a step back from composing music and see if we could design the artists themselves.

 

When the bots source other artists recordings, are the bots relying on particular hashtags?

 

Yes, our Soundcloud bots can search for tracks based on things like                                                                                                                                                                                                                                                                    

                                                                                                                                                                                                                                                                                                                               

                                                                                                                                                                                                                                                                                                                               

                                                                                                                                                                                                                                                                                                                               

You have read a selection from Issue 2: Summer-Fall 2015. To read this text in full, purchase a limited-edition print issue in our store for $10.00 (+ shipping) or visit one of our stockists, or download our free reader-style app from iTunes to purchase a digital edition to read on your iPhone, iPad or iPod Touch for $5.00. Annual subscriptions to the digital edition are also available for $10.00.