Last week Sony revealed that it had created two new pop songs by use of artificial intelligence.  The two songs were created at a Sony CSL Research Laboratory using the company’s “Flow Machines” software which learns musical styles from a catalog of songs and then compiles various combos of style transfer, interaction techniques and optimization.

The French Composer Benoit Carre’ wrote the lyrics to the first song, “Daddy’s Car” and also produced and arranged the track which has been characterized as being in the same style as The Beatles.  The second of the two songs is titled “The Ballad of Mr. Shadow” and is similar in style to Jazz Artists such as Duke Ellington, George Gershwin, Cole Porter, Irving Berlin and others. 

But wait, if Carre’ wrote, produced and arranged the tracks, why is Sony telling the public the music was created by Artificial Intelligence?  Well, Carre’ used a database setup by Sony Called LDSB which contains over 13,000 lead sheets from a lot of different composers and styles.  Carre’ just selected the style and came up with a lead sheet using the Flow Composer software. Only the lead sheets were computer generated, the “Flow Machines” software generates lead sheets based on a specific style which then are matched up with other audio components, including lyrics which the composer wrote before the tracks were completed.  By the use of the Rechord systems, Carre matched pieces of audio with the generated lead sheets, then all he had to do was finish the mixing and finishing touches of production before the final cuts were ready.    

Both tracks can be listened to online for free at https://www.youtube.com/watch?v=LSHZ_b05W7o and https://www.youtube.com/watch?v=lcGYEXJqun8, to Carre’s and Sony’s credit both tracks sound like The Beatles and Classic Jazz musicians, although choppier and perhaps more like a random B-side track that you never listen to.  Sony has revealed that it will release a full album of music generated by Artificial Intelligence in 2017, however at this point the music is only partially generated by AI and AI software and will contain audio components created by humans and pieced together by a composer.