Over the weekend I read this fascinating article in The New York Times Sunday Magazine about the predictor algorithm that Netflix has been working on for years. The online movie rental giant has put forth a contest challenging people to improve its recommendation system. Anyone who can increase its accuracy by 10 percent will receive $1 million.
If you have a Netflix subscription, you’ve probably noticed that the company harasses you about rating every movie. The reasoning goes that if you keep rating movies, Netflix might be able to recommend new ones that you like. It’s similar to the technology that Amazon and iTunes use to recommend books and music you might like.
The major obstacle to improving Netflix, it turns out, is quirky fare like Napolean Dynamite:
Worse, close friends who normally share similar film aesthetics often heatedly disagree about whether “Napoleon Dynamite” is a masterpiece or an annoying bit of hipster self-indulgence. When [51-year-old “semiretired” computer scientist Len] Bertoni saw the movie himself with a group of friends, they argued for hours over it. “Half of them loved it, and half of them hated it,” he told me. “And they couldn’t really say why. It’s just a difficult movie.”
Mathematically speaking, “Napoleon Dynamite” is a very significant problem for the Netflix Prize. Amazingly, Bertoni has deduced that this single movie is causing 15 percent of his remaining error rate; or to put it another way, if Bertoni could anticipate whether you’d like “Napoleon Dynamite” as accurately as he can for other movies, this feat alone would bring him 15 percent of the way to winning the $1 million prize. And while “Napoleon Dynamite” is the worst culprit, it isn’t the only troublemaker. A small subset of other titles have caused almost as much bedevilment among the Netflix Prize competitors. When Bertoni showed me a list of his 25 most-difficult-to-predict movies, I noticed they were all similar in some way to “Napoleon Dynamite” — culturally or politically polarizing and hard to classify, including “I Heart Huckabees,” “Lost in Translation,” “Fahrenheit 9/11,” “The Life Aquatic With Steve Zissou,” “Kill Bill: Volume 1” and “Sideways.”
In other words, the programmers tackling the Neflix problem are attempting to tackle whether human taste can ever be totally predictable based on past behaviors. Quirky comedies and outrageously political commentaries seem to be hard to predic –the more controversial the film, the harder it is to say if someone will like it or hate it.
The interesting thing about these programs it it sets it up as a computer versus human scenario. But the programmers are examining very human questions. For instance, Bertoni realized the difficulties posed to Netflix by widely panned films that fit into a user’s preferred genre. As long as humans are addressing those questions on the programming end, it will be a combination of computer and human prediction.
Cross posted at Pushback.
No comments:
Post a Comment