When Netflix offered $1 million to anyone who could help them suggest movies better, thousands of teams from hundreds of countries signed up for the challenge. Netflix uses a program called Cinematch that recommends movies to its customers, designed to keep the customers renting movies and paying money. If people could create a program that would suggest movies 10 percent better than Cinematch, that team would win $1 million from Netflix.
One team at AT&T Labs came particularly close to that goal and wrote about the competition for the latest issue of IEEE Spectrum. The team members combined a number of different search methods to create a program that was 8.43 percent better than Netflix’s. That’s wasn’t enough to win the $1 million dollar prize, but Netflix was also offering a $50,000 prize to the team that came the closest.
Programs like these are capable of “finding something out about us that we ourselves can’t even figure out,” writer Clive Thomas told the WNYC show On the Media. They also run the chance of perpetuating narrow-mindedness by suggesting only media that people are sure to like, without any of the mind-expanding media that people might aren’t sure to enjoy. People’s friends, rather than computers, are still better able to suggest media that might not be as enjoyable, but is still important.
Computers may be able to explore the “impenetrable mystery at the heart of our predilections,” according to On the Media’s Brooke Gladstone, but they aren’t able to change those predilections without the help of a few friends.
You can listen to that interview below:
Image by Urthstripe, licensed under Creative Commons.
Sources: IEEE Spectrum, On the Media
