Recently I ran into this article by Malcolm Gladwell (The author of The Tipping Point and Blink) about people writing software that analyze songs and movie screenplays, and predict which ones will be hits. In fact, one company, Epigog, uses a neural network to analyze screenplays, based on the connectionist theory I'm on about so much. Apparently, Epigog's neural network accurately predicted the box office grosses of several big movies within a few million dollars. However, they're now employed by a major movie studio, and their technology has become fairly hush-hush.
More observable is Platinum Blue , software that analyzes music looking for songs that could chart. Some excerpts from the Gladwell article:
Four years ago, when [company owner] McCready was working with a similar version of the program at a firm in Barcelona, he ran thirty just-released albums, chosen at random, through his system. One stood out. The computer said that nine of the fourteen songs on the album had clear hit potential—which was unheard of. Nobody in his group knew much about the artist or had even listened to the record before, but the numbers said the album was going to be big, and McCready and his crew were of the belief that numbers do not lie. "Right around that time, a local newspaper came by and asked us what we were doing," McCready said. "We explained the hit-prediction thing, and that we were really turned on to a record by this artist called Norah Jones." The record was "Come Away with Me." It went on to sell twenty million copies and win eight Grammy awards.
Also-
This past spring, for instance, he analyzed "Crazy," by Gnarls Barkley. The computer calculated, first of all, the song's Hit Grade—that is, how close it was to the center of any of those sixty hit clusters. Its Hit Grade was 755, on a scale where anything above 700 is exceptional. The computer also found that "Crazy" belonged to the same hit cluster as Dido's "Thank You," James Blunt's "You're Beautiful," and Ashanti's "Baby," as well as older hits like "Let Me Be There," by Olivia Newton-John, and "One Sweet Day," by Mariah Carey, so that listeners who liked any of those songs would probably like "Crazy," too. Finally, the computer gave "Crazy" a Periodicity Grade—which refers to the fact that, at any given time, only twelve to fifteen hit clusters are "active," because from month to month the particular mathematical patterns that excite music listeners will shift around. "Crazy" 's periodicity score was 658—which suggested a very good fit with current tastes. The data said, in other words, that "Crazy" was almost certainly going to be huge—and, sure enough, it was.
I looked around and found this interview with the company owner on BBC radio, which actually plays a bit of music that their software predicted would be a hit, and found this song by an unsigned artist on myspace that is supposed to rank as a potential hit.
Understandably, a lot of people are unhappy about this. Here's a link to another blog that sees this development as us getting closer to machines dictating to us what sounds good, and hypothesizes that if a machine had delivered a mathematical judgment on Elvis or the Beatles, they may never have gone on to make some of the greatest pop music of all time.
Listening to the hits of the system predicted, it does seem to me that the software really is latching on to some quality of popular songs. The unsigned bands it hails differ a lot from Gnarls Barkley and Norah Jones in terms of style, but they share a similar sonic quality. Lots of reverb, with the notes leaking on to one another. And while its hard for a mere human to describe in words, The melodies all have a similar lilting, comforting quality.
Those qualities may not be the only things that matter in music. I actually don't always prefer these types of songs, even if I acknowledge that the artists that deliver this type of sound always seem to become more popular than the ones I really love. But on a basic level, if you want to find a song that will register on a similar level to James Blunt's "You're Beautiful" or Dido's "Thank You", it really does seem that it could help find those tunes.
I have two problems with it though-
1. Its predictions are tied to the past, not the future.
The software works by comparing the mathematical formula of new tracks to the formulas of thousands of billboard charting hit songs (and thousands more non-billboard charting non-hits), to see which patterns the new song are closer to.
Norah Jones, James Blunt and the song "Crazy" are all quite good. They're also quite conventional, and don't particularly differ from anything that was popular 30 years ago. (With their bizarre stage presence, Gnarls Barkley may seem to break this mold. But actually, the music is sampled from an old song from the 60's, and some have attributed its success to its classic qualities)
For now, that pattern works great. However, if this software became standardized, and trial and error became completely eliminated, no radically different music would ever get a shot. Then the "hit clusters" that the software tracks would stagnate to what had gained popularity before the software became standard. And we'd be stuck listening to what was a considered a hit up until 2007, with no further variation.
Would this software have stopped the fathers of rock and roll from getting big, just because they sounded sonically dissimilar to Frank Sinatra or Bing Crosby, or whatever other crooner was popular before they came along? What about Little Richard? What about Chuck Berry? What about those thumping beats? I don't know much about how music looks on a spectrograph, but I suspect that those spiking metronomes would have made those songs look extremely different from hits of years past. It sure as hell wouldn't look like "You're Beautiful" by James Blunt. Sure, the software would match songs by those artists to one of its 60 "hit clusters"...now. But people have to make the judgments on new hit clusters in the first place. I suspect that in the recent past, some of those hit clusters didn't exist.
What about techno? What about Missy Elliot's "Pass that Dutch", a hit from years ago that used a digeridoo and layers of handclaps for its rhythm section? I don't doubt it shared a lot of qualities with hits of the past...but it also had a lot of very different qualities that likely would have sabotaged its Platinum Blue rating, at least at the time.
What about rap? Would it have liked Run DMC's first album back in the early 80's? What about techno? Would it have forecasted the success of The Prodigy's "Firestarter" -a massive worldwide hit consisting of electronic squeals and a ranting punk- back in the late 90's?
Now, you may hate techno and rap, and see that as a sign that the "James Blunt-ness meter" is doing its job. But that brings me to my next point.
2. There's no accounting for taste, and some people's tastes aren't profitable.
Actually, this reminds me of another Gladwell article, about how food companies started to look for one perfect flavor for a product, because one that pleases everyone equally often turns out to be pretty bland.
So instead of settling for one spaghetti sauce, that got the best overall average score, say, 6.5/10, they started putting out 3 or more flavors that each got scores of, say, 8/10 with a third of the population, even if each flavor's overall numbers were dragged down in the overall average by other people that hated it just as strongly. That's why, say, extra spicy chips are profitable. If it was the only flavor, 70% of the population would avoid it altogether. But the other 30% of the population prefer it over all else.
Apparently, the same thing goes for music. If you listen to the BBC radio broadcast above, someone designs the worst possible song by doing a profile of things about music everyone hates. And yet, and yet...he goes on to say that there are still several hundred people that prefer that particular type of song to all else. They might be the smallest group of all. But they're still out there.
And here's the thing- if your group is small enough, you might find that companies don't want to cater to you. Even the 3-30 major flavors of a given spaghettis sauce remain, to some extent, compromises between what you and millions of other people want. For example, my favorite pizza has lebanese donair kebab meat, feta cheese, banana peppers, black olives, onions and mushrooms. But I don't expect to see it sold pre-made at the supermarket anytime soon. To make everyone as happy as possible, Companies would have to put out hundreds of variations of their products, and that just isn't profitable on a mass scale.
In music, however, we already have those hundreds of variations. Tens of thousands actually, if only by trial error. So while the software may make the lowest common denominators easiest to strike on, and make the business more profitable, if you're a music fan that listens to 5 or more new albums a month, this kind of standardization could actually hurt you.
I listen to a lot of mainstream stuff too, but one of my favorite CDs a few years ago was by a failed white rapper called Cage that rapped about getting beaten by his biological father as a child over acoustic guitars. No-one else I've played it to likes it much. If Platinum Blue analyzed it, it would almost certainly give it the thumbs down. (And rightfully so, at least from a profit point of view -It didn't sell jack). But I would be a lot worse off if the machine had made the decision to cut it. And your own personal Cage could get the cut too.