modern recommendation engines like Netflix ... the system does such a fine job of finding you something to watch.
I strongly disagree (about Netflix). I think their recommendation engine is dismal. It pretty much never finds me things that I'm interested in.
To me, the strongest evidence that Netflix is after a red herring is the Netflix Challenge. The goal was an algorithm to provide the most accurate ratings, but I think that phrasing the question in that way makes it almost irrelevant.
The real question should be: what movies will this user likely find to be very good. That means that a precise rating is never necessary, and no rating is needed at all for things that are outside the 4-5 range. And an algorithm to find movies that simply have a high likelihood of high scores is a very different beast from an algorithm that can precisely rate movies across the whole spectrum.
modern recommendation engines like Netflix ... the system does such a fine job of finding you something to watch.
I strongly disagree (about Netflix). I think their recommendation engine is dismal. It pretty much never finds me things that I'm interested in.
To me, the strongest evidence that Netflix is after a red herring is the Netflix Challenge. The goal was an algorithm to provide the most accurate ratings, but I think that phrasing the question in that way makes it almost irrelevant.
The real question should be: what movies will this user likely find to be very good. That means that a precise rating is never necessary, and no rating is needed at all for things that are outside the 4-5 range. And an algorithm to find movies that simply have a high likelihood of high scores is a very different beast from an algorithm that can precisely rate movies across the whole spectrum.