Artificial intelligence is earning itself a seat in just about every faculty; it’s in mental health, it’s in productivity and office assistance, it’s in web design, it’s in fashion, and it’s in legal research. Now, it is smack dab in the middle of music.
Last year we told you about AI systems that could automatically compose instrumental songs based on an understanding of moods, genres, and some “basic building blocks.” Now, welcome IBM Watson to the show; it’s collaborating with Grammy-winning producers and pop stars. Specifically, Watson’s music algorithm, Beat is helping fill the lyric books and ledger lines.
To provide ample writing assistance to producer, Alex Da Kid, Watson’s research was extensive; it studied tens of thousands of songs to pick up trends in structures, key signatures, melodies and even note velocity (how hard a note is struck); it reviewed everything from speeches to New York Times articles to uncover themes in content commonly capturing the public’s attention; then it reviewed the social activity corresponding to all that content so that it could map themes to sentiments and, as IBM puts it, “build the ’emotional fingerprint’ of each of the last five years.”
From all that, Watson pumped out scores, song snippets, bass lines, lyrics, and other components—it even created an interactive visualization of the data based on palette of preferred images, colors and album artwork. Alex Da Kid then took over, absorbed inspiration, took what he liked, discarded what he didn’t, added his own, and ultimately produced the song below, Not Easy. Don’t worry, it’s only the first song of a 4-song EP Alex, alongside Watson, plans to release.
Enjoy, SnapMunkers: you’re listening to the future. It sounds a lot like a Budweiser Superbowl commercial.
Also, here is some “behind the scenes” footage that provides some more background and context to the process and final product.
SnapMunk
It seems important to clarify that the words intelligence, creativity and cognitive are being used metaphorically in computer science.
AI generatively recycles and recombines massive amounts of data provided by humans, and this is what gives it an appearance similar to individual and collective intelligence. In a related way, the expanding use of audio sampling/looping in music provides a digital simile of physical instruments and musicians only because its database approximates, recycles and recombines the prior recordings of human musicians.
In either instance, these similies of human intelligence and creativity are dependent on massive quantities of encoded, discrete, snippets of human content (and human generated algorithmic recombination patterns) for their approximation. Thus words like intelligence and creativity are being used metaphorically when characterizing AI as being anything more than an appearance/simulation of the qualities of consciousness and sentience.
Current trends in technology and AI are replacing an increasing number of human activities. In this context, it seems important to distinguish the word ‘creativity’ in (at least) two senses: creative process (composer) versus generative recombination (compositor).
Ultimately, it seems imperative to value and develop human creativity as a process to which new technology is designed to assist; as an area distinct from training the human creative process for the role of assisting (editing, arranging) a generative technological process.
Rad. A computer that can create a song much better than I can. Great!
At this point, I am quite impressed with AI created music. It really doesn’t sound that much different from other electronic music that we hear today.
Maybe Watson can perform at the Super Bowl half time show?
I will +1 this for sure!
Very true. The half time show at the Super Bowl is quite lame. Maybe something like this would be more entertaining for those that are getting up for more beer, nachos or to use the bathroom.
Awesome, I would love to see more information about the commonalities between songs these days.