CMU School of Drama


Wednesday, October 26, 2016

Sony is developing an AI that can create new music from existing songs

Business Insider: Sony is developing an artificially intelligent system that can create music from data in existing songs, Reuters reports.

The system, which is called Flow Machines, could expedite the music creation process, and have applications in areas ranging from audio and video production to advertising.

2 comments:

Chris Calder said...

The idea of having a machine write music for you is definitely hard to grasp. I can see people having many problems with this product. First of all, I highly doubt artists will like a computer manipulating the art that they created and I certainly can’t see people getting an emotional reaction out of a piece that was conjured up by a computer. People listen to music because it is something that they can relate to. Knowing that the words were written by a person allows us to connect to the music on another level. If you take that element away I can't imagine you will have people banging at your door to listen. At the end of the day I wouldn't be surprised if I saw this AI being quickly dissolved simply because of the negative press that it will receive, but hey maybe I will be wrong and this will be Sony’s major comeback. In any event, the way in which a consumer views art is forever changing and only time will tell what is next for this industry.

Unknown said...

After watching a lot of Black Mirror this week, seeing this article doesn't surprise me. I guess its disconcerting that a robot is taking an art form and creating it artificially, but in a way, it kind of makes sense. We, as mere humans, cannot possible remember every single song that's ever been produced. That's why we hear about copyright infringement lawsuits in the music industry all the time. Having a literal hard drive filled with every song imaginable would cut down on that considerably. However, how do you quantify music to a robot? Yes, you can give them the notes on a piece of paper, but so much of a song is the emotion in which the performers play their instruments (or sing). Different timbres of voices, how one song flows into the next? Forget about a robot actually writing these songs, how does it recognize good music over bad music?