(no subject)
Nov. 24th, 2008 11:56 amGotta love this article about the competition to improve Netflix's viewing suggestions. I, for one, think $1M is totally appropriate, if not even a little lean of a prize to offer. I've personally reviewed at least 1500 titles on Netflix only to get about one or two recommendations, usually out of the blue, no-way-I-would-ever-want-to-watch-that types. Sure, when you add something to the queue, you get a pop-up window with about ten new selections, but they're tiredly predictable. (Just because I wanted to see North by Northwest doesn't mean I have to see everything with Cary Grant or directed by Alfred Hitchcock nownownow.)
This is all relevant, too, seeing as the new Netflix-enabled XBOX 360 upgrade is, in a word, awesome. The picture quality is superb, and it automatically upgrades if you have the capability to broadcast the feed in HD. (We do, and we did.) It takes hardly any time at all to download and get started. (At least, one episode of a TV show did; I have not tried any movies just yet.) It works beautifully. Yes, you have to have a gold account to get the streaming across, but
feiran does, and it didn't ask us to prove that she is in any way invested in my Netflix account. (The set up process took all of a minute. Bonus!)
And, relevant to this development: cinema is dead? (Again?) Yes, let us play the dirge for the cinema, it's gone gone gone. Because it's not like teenage girls just dropped $70M on Twililght this weekend or anything...
This is all relevant, too, seeing as the new Netflix-enabled XBOX 360 upgrade is, in a word, awesome. The picture quality is superb, and it automatically upgrades if you have the capability to broadcast the feed in HD. (We do, and we did.) It takes hardly any time at all to download and get started. (At least, one episode of a TV show did; I have not tried any movies just yet.) It works beautifully. Yes, you have to have a gold account to get the streaming across, but
And, relevant to this development: cinema is dead? (Again?) Yes, let us play the dirge for the cinema, it's gone gone gone. Because it's not like teenage girls just dropped $70M on Twililght this weekend or anything...
no subject
Date: 2008-11-24 08:50 pm (UTC)Increased site access would be fabulous. I was having to add movies from my PC to the XBOX, which, for me, isn't hard since I have a laptop. But that was still an extra step that really could be streamlined, especially for people without laptops. I'm not as phone-tech enabled, but I can see how mobile service is a good idea, too.
The movie recommendations need to request a lot more specific information from customers, though not with the biological/geographic data (as the article specified), if they're going to work better. My feeling is that you could make it better if there were another option when you pick out a title to say why you were picking it. It's another layer and most people wouldn't use it. But for the people genuinely interested in getting new suggestions, it would be really helpful, and it's just one more rating tick. Like, if you threw on their example of Napoleon Dynamite, say, because people wouldn't stop talking about it ("Friend recommended") or you read a review about a movie you were interested in but not going to pay to see in a theater, that would help define user tastes much better. But, like I said, it's more work for the users, so it might not necessarily fly.
no subject
Date: 2008-11-24 10:31 pm (UTC)Mobile access is a little silly, I admit, as it is for many sites for which it's available. But for someone like me who hasn't seen *anything* and has a mind like a sieve, the idea of being out and hearing someone mention a movie and being able to add it then and there (which, admittedly, I can do now with my DVD queue, just not instant) is nice.
If they do move to a "singular value decomposition" system as mentioned in the article (“You can find things like ‘People who like action movies, but only if there’s a lot of explosions, and not if there’s a lot of blood. And maybe they don’t like profanity,’”) they definitely should do as you say and allow users to tweak those parameters themselves. The number of possibilities would be massive, and in the end it'd have only one's own ratings to go on. I've rated around 280 movies, and I feel like that's probably not enough to nail down even some tastes I'm very consciously aware of.
But that's just one of countless potential pitfalls in a system like this. I usually do a pretty good job of guessing from a trailer whether I'll like a movie or not, so I have lots of 3-5 star ratings, but very few 1-2 star ratings. While it may have a good grasp on what I like, it probably has no idea what I hate, and even the best system can only go on the info its given, so my distaste for gory movies (while enjoying action and more psychological horror) may not be a parameter it can nail down without my help.
No system will ever be perfect in this regard, anyway. My closest friends don't always guess right what I'll like, nor do I myself (I said I'm "pretty good" at it). Personal tastes may be sufficiently complex to amount to a chaotic system which can't be modeled VERY accurately no matter how much data one has.
Expectations should be managed, too. I also sometimes see a trailer or read a review and am entirely uncertain if I'll like the movie or not. If the system gave not just a predicted star rating but a certainty level, it'd be more meaningful. They could either show the certainty to users to choose as they like, or if they thought that too complex, just let people select an option like "Show me anything you think I might like" versus "Show me only things you're pretty sure I'll like" (maybe make showing the certainty value itself a third option).
Makes me wistful for my days on the Artificial Intelligence track in Columbia's CS department....
no subject
Date: 2008-11-25 02:45 am (UTC)