Jul 23, 2009
There is a future waiting for us -- a future that's not so far off -- in which we won't bother to put filmmakers' names on their films. Musicians' names aren't going to appear in online music stores' listings. We might still search for books to buy, but we won't search by author. Admittedly, this can't help but seem contrived and far-fetched. Creative people demand credit for their works, don't they? If their labors aren't rewarded well financially (and often, they aren't), then they at least need to be recognized. Right?
Well, I should be more specific. It's not that authors, bloggers, filmmakers, and musicians are going to agree collectively to have their names scrubbed from their works. Rather, it's that nobody actually searching for and using that content is going to care who made it. Amazon and iTunes will drop the byline column in their interfaces. In fact, our online shopping won't resemble iTunes or Amazon at all. Life is moving swiftly away from that clunky, old-school model.
Before I explain my vision, let me give some background. There are two main reasons that the concept of authorship is dying out. First, technological progress in the realm of communications -- particularly internet communications and wireless personal devices -- have made it possible for the Joe Six-packs of the world to create and publish content til they're red in the face. For what it's worth, a recent Wired article makes a number of important points about the consequences of this new "everyone's a director" era. Not the least of these, of course, is the sheer quantity of potentially interesting content we have on our hands now. But it's also worth noting that allowing everyone to be a journalist and a movie maker means that each of us is going to find more content that resonates with us personally than ever before.
Second, the technology we use to comb through the reams of (mostly useless) content now confronting us has progressed immensely over the last couple of years. We have a two-pronged strategy, and it seems to be working. On the one hand, better robots: software engineers are always offering up improved algorithms that learn more about what we're looking for when we search for certain terms or navigate certain sites. On the other hand, more humans: the effort to manually catalog all Earthly data -- embodied by Wikipedia, the Human Genome Project, the Music Genome Project, and many others -- have made huge strides as well. When you allow everyone to contribute just a little bit, you wind up with a whole lot of collective elbow grease.
What's in a name, anyway? Decisions based on the reputations of others have been a staple of business and politics for as long as humans have had tongues to gossip with. Your name is a placeholder for the relevant, public information others may need to know about you -- in short, what you're 'worth' to the tribe. Once upon a time, societies were small enough that everyone was fairly well-acquainted with one another. No longer, unfortunately. Once we began settling down in bigger groups, it was suddenly important to have ways of predicting the tendencies of other folks in the tribe. Suddenly, we found ourselves in need of heuristics (academic-scientific jargon for 'shortcuts').
Luckily, human beings generally abide by the law of momentum. Wild swings in decision making, personal preferences, and habits are likely to be taken as manifestations of psychological disorder, precisely because they're fairly rare. As such, my reputation (or name) is a reasonably trustworthy indicator of what others can expect of me when they encounter me face-to-face.
But we're talking about searching and filtering content. How is the whole subject of personal reputation relevant to our discussion? As it turns out, the reputation dynamic applies just as much to people trading in a marketplace or negotiating contracts as it does to the production of literature, painting, sculpture, music, theater, journalistic content, and other media. Unless you're with the artist in her studio, you have no idea what's around the corner. If you're a fan of the last few Animal Collective albums or David Sedaris books or Nick Kristof articles, there's a pretty decent chance you'll like whatever they put out next (or, for that matter, what they put out a long time ago).
But what do you get when you combine an exponential increase in the amount of content, a democratization of that content (so that most of it is made by "nobodies"), and the ability to sift through all of that content more accurately than ever before? Names are no longer the best available heuristic. Plenty of nameless nobodies create great content. Searching for the big names won't help you find them. At the same time, the big names don't always offer what you're looking for. The powers of great artists and writers often decline. Consequently, our methods can (and must) evolve. And they've begun to. I can type "Animal Collective" into Pandora and gain instantaneous access to untold hours of music by other artists that may suit my mood better than whatever I might find by simply searching iTunes for more of the Collective's own releases.
What would the world look like if we applied the Pandora model to realms other than music? Let's take news. For generations, people have relied on knowledge, built up over years of browsing different sources, that certain publications and authors are particularly reliable and relevant. These publications and authors produce trustworthy, well-written content. Further, each of us has a mental list of the periodicals that sync well with our personal interests -- and these are the ones we find ourselves visiting habitually. But let's imagine instead that we can break these routines we fall into. We have good reason to. After all, I might be a fan of, say, the Washington Post, but it's fairly commonplace that I can't find much of interest past the front page.
Instead, what if we take all the electronic news produced around the world and apply a set of filters and machine learning algorithms to it? Our filters produce a newspaper's worth of articles, plucked incredibly selectively from all that data, delivered to each of our inboxes each morning. Google News is one of the first baby-steps in this direction, but there's so much more to be done. It does a fairly good job lumping together related content from multiple sources, but it can't learn a thing about your interests and tastes.
But what if, as people read articles, they were able to select several attributes from a finite list (not free-tagging, mind you) that they thought applied to that content? Then, a machine learning algorithm would discover similarities and patterns among content with particular combinations of those attributes. For example, a set of similar articles might be tagged with "North Korea," "analysis piece," "over 1200 words," and "small news outlet."
With a system like this in place, you and I would be able to cruise from article to article based on those underlying patterns. With a combination of user-set preferences ("I'm interested in articles about international affairs and the environment") and machine-learned preferences ("This user mainly likes analysis pieces that are over 1200 words"), the software could deliver a highly-tailored collection of articles every day, just as Pandora delivers streams of music customized to individuals' tastes. Of course, machine learning does best when it has lots of user input, and crowd sourcing for that input is a quick and easy way of getting it. Like I mentioned earlier, a little elbow grease from a lot of people translates to a lot of elbow grease. This is something we could begin building today.
An album used to be a fantastic way to structure an hour's worth of music. It was all written by the same person or people over a fairly discrete period of time. Songs arranged into an album often flow well into one another, and the timbre of an artist's discography usually shifts slowly over time. In the same way, a newspaper managed by a unified editorial staff used to be the most logical way to organize journalistic content. The politics and perspectives of such a staff are often slow to change. End users are slow to change, and so attaching themselves to slowly-changing entities like bands, writers, and newspapers made a lot of sense. But we're entering an age in which it's no longer necessary to rely on the 'human momentum' heuristic. We have new ways of grouping content -- that is, we can collect things under categories that are guaranteed to change only when the users themselves change their minds. After all, momentum is only steady until it shifts. If we have better ways to find and group content, let's use them.