OK computer: When pop music meets machine learning
It’s Moogfest season here in Durham, so there’s been a lot of the discussion in the office around music, data lakes, and the heat map we’re building for the festival. But the conversation took a different turn, thanks to a tweet.
Many months ago when I was at IBM Insight, I tweeted a snide remark about computer-generated jokes. Fast-forward to this week, when former “Monk” and Letterman writer Joe Toplyn responded with a link “proving” that computers could generate jokes that were funny … at least to the easily amused. Amid the discussion, someone drove by playing crappy autotune pop music.
This got me thinking about whether you could generate hit pop songs. Most of the popular songs are written by two middle-aged guys from Sweden anyhow. Plus, there are algorithms that can detect which songs are likely to be a hit. While the current hit song generator is simply song titles with performers, we also have an algorithm that can generate tweets for the presumptive Republican presidential nominee. It seems like a short trip to get from hit detector to factory songwriting to neural net for political speech to full-on pop song generator!
We’d need parameters like a genre (pop, hip-hop, dance) and probably gender, as well as whether it’s a party track, a love song, happy, sad, angry, and so on. Then maybe we’d train a neural net on the corpus of songs by the two Swedes. Add that to an adaptation of the hit detection algorithm and you should have not a great song, but at the very least a popular one.
Unlike the acts at Moogfest, which tend to be more complex, modern pop music is especially appropriate to this approach. Like a business, it has a limited grammar or topic area. The beat and the musical accompaniment fall along more predictable lines. and there is an almost cyclical behavior to it.
As in business, should we achieve this level of automation, we could even stop doing chart toppers and instead make a custom hit for each customer. By profiling your reactions and moods, we could play what you most want to hear right now: a fresh (but predictable) tune out of our Spotify, composed expressly for you in real time. As in business, we’d save cost by stratifying this function and identifying like music listeners, then composing a smaller set of tunes on the fly and picking between them.
As with many companies, a lot of manual labor goes into even a factory composition and certainly in the performance. Pop stars effectively run a cottage industry of marketing the songs. Though we could automate the role of the two Swedish composers, it would be more difficult to automate the pop stars’ work. Granted, many of these songs aren’t difficult to perform, so a group could show up and sing Lady Ga Ga’s “Telephone” in a bar with no previous rehearsal. However, the human element of the performance and the creative that goes into the branding and marketing are harder to replicate. The younger generation may become ready to watch S1m0ne perform it on Periscope, or maybe the band of the future looks more like Gorillaz.
Unlike in business, machines could probably generate the songwriting without anyone noticing, and we don’t have to push for people to just let go to gain mass adoption. Many pop fans already believe the performers write the tunes. Wouldn’t that be a kick if we achieve mass adoption of machine learning in songwriting before we achieve full data-driven decision-making in business?
Source: InfoWorld Big Data
