Hacker News Reader

Comment by jmtame | Hacker News Reader

2 years in and they still haven't given out the prize money. How's that crowd sourcing working out for you NetFlix?

A 9% improvement represents a substantial increase in quality of recommendations. If I remember correctly, a 10% improvement meant the difference between a complete star rating, on average. (There was a fairly substantial write-up about what the 10% meant last year.) I'd say for the people-hours put in to the project, they've gotten more than their money's worth.

Oh I bet they have. Have you calculated the number of hours the 35,000 teams put in relative to the amount of pay they received?

I bet they've gotten more than their money's worth.

At what rate should we value their hours? Do you charge more or less for doing something you like?

Have you considered that voluntary trade is not zero sum?

Voluntary as in they don't get paid?

Because I'm sure that every single person has no intention of receiving any of the prize money. None whatsoever, it's purely voluntary, and nobody deserves any money for making Netflix more profitable.

Wow you guys really want to bring it out of me on this thread. That's alright though, I dig everyone's energy (and I don't mind taking a bit of a karma beating over this either).

Actually, at the time of writing this, as a result of my discussion, other people have been awarded 28 karma points (where I have suffered a loss of a few). Outside discussion accounts for 35 karma points. So relative to the overall karma, I'm helping a lot of people out by being the grumpy, grouchy, mean old Mr. Scrooge in defense of the researcher. I do it for you guys! :D

You don't think the people entering this competition can at the very least calculate the odds of winning?

Now if only it was possible to blacklist trolls...

I don't think you can silence people who disagree with your views on HN, but goodluck trying.

Hey, it is not easy to develop a better recommendation engine. They have problems like the "Napoleon Dynamic" which is still remained unsolved right now.

Computer can only predict human behavior which is unchanging and static. But, this is always not the case.

I didn't say it was easy, I'm not sure where you're drawing that assumption from.

I did essentially say that over 2 years, I'm glad that the collective research efforts of 35,000 teams, one of the teams has seen $100k from a corporation who will ultimately profit off their research.

One of the teams mentioned in the article started they own company based on the notoriety achieved, so yeah. It seems to be working pretty well for everyone.

By the way, your post was just crawled by Google, and it probably represents a couple cents into their billion dollar quarter. How do you feel helping the big guys, huh? Working for free for the man... you're disgusting, sir!

Yes, I've been bad! I'm going to sit in time out bbs

LOL. I misunderstood what you have said. Since the code will be open sourced, it will definitely benefit thousands of companies.

In fact, there was one graduate student who was in our campus last semester was doing research in this area. He did a very good job on this. And, he won the first ACM machine learning prize for the paper. If you are interested in this area, you can actually go to the library and find out the paper. Now, I think he is now working in the Microsoft Research Team(the usual career path in our campus).

I actually did a tiny recommendation engine using one of the collaborative filtering algorithm - slope one predictor last two days. It will be up on the web very soon. I am having problems updating my server because rubygems.org is down!

It's all good, I think I need to stop hanging out on YC today. I have to catch up to Matt on this Cocoa Programming book, and then I've got a final project to prep for, and finally a meeting tomorrow with the company to talk about some marketing stuff.

Clearly YC is not where I should be right now :)

You are wrong that it can only predict static and unchanging behaviour. It might just be that there isn't enough data to draw any relevant conclusions about Napoleon dynamite.

Terrible! We so badly want to give away a million dollars, and yet all these people will do is give us a massive improvement on recommendations without taking our money! We are pissed

- Netflix

Successfully giving out the money--especially if there is (for example) the romantic, swooping in ending that staunch stated that he wanted (in his post above)--will generate a lot more positive PR for Netflix than the remaining 0.44% improvement will help them at this point. Regardless of whether you think of this as a genuine research prize, or simply a corporate ploy, you have to admit that Netflix is probably genuinely unhappy that no one has managed to complete the competition yet.

It's working great. In fact, I'd say it's fucking awesome. The recommendation algorithm has improved greatly, and contributions to the science of statistics have been made because of this corporation.

Your comment strikes me as obnoxious. You try to gibe NetFlix, but the truth of the matter is that there is a market. The considerable prize sum is enough to interest a competent individual.

Sorry, you'll have to excuse that I'm not immediately a huge fan of corporations who give small amounts of money for significant amounts of research that ultimately go back to benefit the corporation.

Apparently, they're going to give all of the code out, which is good to hear. I'm proud of the machine learning community for that. As for Netflix, it's difficult for me to believe this is really about research and not about increasing their profitability.

I could be way off. Maybe $1 million over the course of 2+ years is a heck of a lot of money for a team of developers (selected from 35,000 teams), which will be used to generate more profit for NetFlix. Who knows. I'm glad that so far over 2 years, $100k was given out. That's good to see.

As for Netflix, it's difficult for me to believe this is really about research and not about increasing their profitability.

Are you really more concerned with the purity of their motives that the results of their actions? I would rather interact with people who can accomplish stuff for selfish reasons than people who fail hard but feel good for trying.

I'm not concerned, I just said it's great that after 2 years of work, we've seen $100k come from it. That is wonderful news.

Why is this downmodded? I can understand people disagreeing, but it is a well articulated point.

Colorless green ideas sleep furiously.

I downvoted after reading the first sentence.

Sorry, you'll have to excuse that I'm not immediately a huge fan of corporations who give small amounts of money for significant amounts of research that ultimately go back to benefit the corporation.

While the statement that followed was at least reasonable, that line struck me as really obnoxious.

You have to remember that many of these researchers are funded by government money (eg, university researchers) or shareholders (eg, corporate labs) that may not be seeing any return on their investment. In some sense, Netflix is stealing cycles that may have been put to better use elsewhere. This could be seen somewhat akin to bribing a public official, only in a 'steal one penny from a million officials' kind of way.

In addition, just think of the massive amounts of cash these researchers are stealing from the network TV execs and advertising agencies, since they may be working on this in their spare time in the evenings instead of watching TV. Despicable.

Funny, but obviously my comment doesn't apply to anyone's spare time.

Reply on news.ycombinator.com