Can Computers Figure Out What Movies We Want to See?

Monday, June 8, 2009

Last week, I read that beeTV, the latest in a series of so-called “recommendation engine” technologies that have emerged lately, announced they had raised eight million dollars. Here’s their video:

I think you’re all aware of the way Netflix uses complex computer algorithms to calculate what movies people like and make suggestions. Although I’m not crazy about the idea of reducing people’s tastes to mathematical formulas and psychological theories, apparently a lot of people find it quite useful. If you’re not a movie buff, but just enjoy movies, how else can you make a choice out of over 100,000 titles?

This kind of technology comes under the rubric of “data mining,” something that makes sense out of tremendous amounts of data, as Wired put it, “trying to make useful sense out of a gigantic dataset, typically rather noisy, completelyNot exactly unintelligible to the naked eye, and, despite its size, often painfully incomplete.” Sounds like an online video service to me. In addition to Netflix and beeTV, other recommendation engines I’ve discovered include WHiWA (“What I Watch”) and jinni.com. (And of course there are things like Pandora in the music world.)

Do these things really work? Netflix has a well-publicized contest going now which is offering a million dollars to anyone who can make their system 10% more effective. While no one has yet been able to reach that goal, some of the entrants have already made the algorithm over 8% more effective. (One of the top competitors for that million is Gavin Potter (aka “just a guy in a garage”), who is now working for….beeTV.)

So it does seem to be effective. Many people are watching the films that the software is recommending. (Of course there may still be a few bugs in the system, as this actual Netflix page above suggests.)

Another aspect of some “recommendation machines” is they allow you to pick a movie based on your mood, as you can see in the beeTV video. jinni-story-tuners-3

Do you want to  see a thriller? How about a sexy thriller? How about a sexy thriller comedy spoof? Jinni.com allows you to search by mood, plot, genres, time/period, etc. It even features “Story Tuners,” which are little sliders where you can fine tune your searches in the various categories, in case you’re in the mood for something approximately 70% Fantastic and 30% Well Known.

All this stuff raises what I think is a very important question: will “recommendation engines” become an ever-increasing part of the way people choose which movie to watch? After all, Rotten Tomatoes already has definitely had that kind of impact. Local critics used to have much more influence in their communities. Now that you mention it, there are a lot fewer formerly influential critics nowadays, largely due to the impact of the internet on newspapers’ bottom line.

Last week, Scott Kirsner asked the readers of his blog, “How Do You Discover Movies?”

“If you think about the last few movies you've seen (whether in theaters, on DVD, via iTunes or BitTorrent), how did you hear about them? Was it via a Netflix suggestion, a Variety review, an e-mail or Tweet from a friend? (Or maybe even an old-school billboard or TV commercial?)”

The large majority of Scott’s commentators made their choices because of recommendations from actual human beings--trusted friends, critics, and bloggers. They watched trailers. They heard about stuff from friends and acquaintances on Twitter and Facebook. They went to films made by their favorite directors and featuring their favorite actors.  And even though they readers of Scott’s blog are likely to be very knowledgeable about film, I think this is the way it is for most people.  But there are people investing millions of dollars in the belief that a lot of people will be open to the idea of letting their computers do the shopping.

These recommendation engines make me more than a little uncomfortable. I’d like to believe that people are more complicated than that. But even if I’m wrong, as my friend John Pierson told me, “it’s a terrible indictment of the state of public herding.”

Pauline Kael Even if people are as predictable as that, movies aren’t. Take a movie like “Anvil.” On the outside it looks like a real-life “Spinal Tap,” and yes, it is funny in that way. Using one of these algorithms you might pass it by, as it’s a genre of one. If you hate heavy metal, it takes a persuasive human being to slice through your resistance and explain why you might like it. That’s the kind of thing that critics like Pauline Kael excelled at. She was a rock star. She led audiences through the power of her writing ability, intellect and persuasive gifts. Her reviews were thrilling events (sometimes more exciting than the movies she was championing). And I’m sorry, Rotten Tomatoes doesn’t fulfill that function. There are many wonderful critics working now, but the great age of the movie critic is over. And that’s a very sad thing.

Maybe this line from the jinni demo video says it all:

“Searching on jinni is like talking to someone who’s really listening to your words.
Sometimes you need a machine for that.”