Hey wiki comics is advertising some pharmacy links you can find below. Sorry for any inconvenience. Hope you can understand... Links are below: Tadalafil Citrate | generic cialis 10 mg | tadalafil citrate 10mg | tadalafil citrate 5mg | generic cialis 40 mg |

Business - Written by on Tuesday, April 22, 2008 20:42 - 0 Comments

Denis Hancock
A deeper look at the NetFlix innovation contest

The first chapter of Wikinomics opens with the story of Rob McEwen and GoldCorp, and it’s a story that Don continues get questioned about fairly often. At the centre of the story was a contest where GoldCorp opened up their proprietary data to see who could answer a couple of “simple” questions – do we have gold, and if so where is it? In the end, the contest allowed GoldCorp to find the uniquely qualified minds that could find gold on their property, and they went from a $90 Million company to a $10 Billion one almost overnight – at the cost of about a million bucks in prizes.

At first blush, it sounds like a great deal for everyone – but these types of contests strike a nerve with a lot of people. Trevor Scholz would be one of them, and Don has a post linking to his article Market Ideology and the Myths of the Web 2.0. To quote Trevor’s take on wikinomics:

When interpreting the new conditions for production and consumption online, his sole mission is a socially friction–free update of the timeworn capitalist power dynamics. He discovers and celebrates how today, perhaps more than ever, fewer and fewer people can become richer and richer by using the very many who earn less and less.

From this perspective, such competitions are disturbing trend – the “capitalists” create a competition, many “laborers” toil away on the problem, and in the end only one or two get paid anything for their efforts, a further extension of the “Winner Takes All” society Frank and Cook wrote about. Not only that, but while the payout to the winner appears relatively large, one could argue it’s a rather small portion of the value that they’ve created.

There are many counter arguments, but the most common is that no one forces people to enter the GoldCorp challenge, or join Innocentive, etc. If someone decides to, clearly they must believe that they are better off for doing so – a.k.a. people make rational decisions in their own self-interest. Responses to this would generally follow the slippery slope principle: while in the beginning this is true, as contests become more and more pervasive, more and more people are forced to join, and stable jobs (and wages) continue to erode away. Someone would then pop in and note that when P&G joined Innocentive their own R&D staff had this fear until they saw that in-house spending actually increased, and it would continue to go around and around from there as people retreat to their ideological corners.

I like to think I lie somewhere in the middle – while I can see the many benefits that would appear to exist within such open competition models and ideagoras, I have blogged about the problems underlying such things from a POV that kind of sits more on Trevor’s side of the fence. But whichever side of the argument one might be on, they’ll probably find some ammunition by looking at the details of the NetFlix innovation contest.

The NetFlix competition is conceptually similar to what GoldCorp did. The recommendation for system is called Cinematch. There are obvious benefits to making it better, so they through open their proprietary data to see if anybody could improve upon it. The competition started in October 2006, and the ”headline prize” is $1 Million, to be given to a person (or group) that improves the prediction accuracy bar by 10% or more. As the 10% was chosen somewhat at random, and they thought it might be a bit difficult and take a while, they also created a $50,ooo “Progress Prize” to be given to the group that made the most (if any) progress in a given year. The rest of the details can be seen in the official rules.

They’re now almost a year and a half into the competition, and we can learn a lot from the leaderboard. As of today (April 22nd 2008), they have recieved 24,923 valid submissions from 3,317 different teams (of 29,587 that have been registered). Right now the pace appears to be about 100 submissions per day. The 10% target has not been achieved yet, so there has been no $1 Million prize – but one Progress Prize has been handed out. Currently, the top performer has a percentage improvement 0f 9.08%, and it appears there are about 50 groups with different models showing 6.65 – 9.08% improvement, so one could seem to reasonably estimate that the goal will eventually be achieved.

But let’s say for a second that 10% is never reached – 9.08% proves to be the maximum. For a cost of about $50,000, NetFlix has harnessed a research group of over 3,000 different teams that have been working on a problem for them that could probably help them make a lot more money. That seems like a heck of a deal (what fraction of an employee could they hire for that amount?), and one can understand why people see a darkside here. In fact, I’m sure that some would argue that even after paying a $1 Million prize NetFlix makes out like  bandit – what’s that, $40 or $50 per submission so far? Anyone care to take a stab at an hourly rate measure on this thing?

But to dig a little deeper into this, one can look at exactly who is doing this work for them – and perhaps learn a little bit about their motivations. Many of them can be linked to from the leaderboard, but here is a sampling that provides some good context  (# represents current ranking):

 1. BellKor team. The current leaders (and the winners of the $50 K prize) are a team of three from AT&T Labs inc. To quote their site: 

We would like to thank Netflix for putting on a heck of a competition. The nature of the competition and especially the collaborative spirit of the participants has made this an exciting and rewarding scientific endeavor. We also would like to thank all of our competitors on the leaderboard for pushing us harder and forcing us to come up with the best possible solution. We look forward to chasing the million bucks with you all!

Well wikinomics me pink – a collaborative spirit among the participants you say?

2. When Gravity and Dinosaurs unite. It seems that two top teams have recently joined together – 4 alumnis from Budapest University of Technology and Economics + 3 alums from Princeton University. Try not to read too much into the ratios :) – here’s a story on them. #11 (Reel Ingenuity) is one of the numerous other academic teams, hailing from the University of Alberta, and #15 from University of Toronto is another one.

3. Team Big Chaos. A team started by two students that continues to grow… an academia trend is emerging, in addition to the team building. For example, #7 Basho added a partner back in September as well. At least one student is working on the NetFlix challenge as part of a course.

6. Acmehill. Well this is a little different: My name is Len Bertoni and I am a member of the team acmehill consisting of myself, my wife, my daughter and my son… My family provides me with new ideas to test out while I subject them to countless “fun hours” of delving into matrix factorization (Dad, what is a matrix anyway?), neural nets (does anyone really know how they work anyway?), etc.

8. Dan Tillberg. A non academic academic – I started working on the Netflix Prize in October 2007 in order to learn more about Machine Learning. I don’t have any degrees of which to speak, but my background is in Physics, Math, and CS, all of which I studied at the University of Massachusetts Amherst. Sounds kind of similar to #9 – just a guy in a garage.

24. An employee of Lockheed Martin, who’s doing it as a project to brush up on his background in machine learning.

27. The Ensemble Experts team: comprised of expert data miners and investment modelers. We are enjoying the challenge of the Netflix Contest and its test of our techniques — forged over the last decade to address real-world challenges.

28. A very interesting one: Mathematical Capital has sponsered a Europe-based team. Mathematical Capital is a new UK venture capital firm specialising in funding early stage mathematical startups.  Our aim is to help those with mathematical ideas move from idea to company.  To do this we offer not only money, but also advice, support and incubation. Think they’re offering money to any of the folks above them in the rankings?

So it seems like we have a collection of professional research labs, professors, hobbyists, and students that have jumped on this as an academic challenge more than anything else. A spirit of collaboration appears to be bubbling to the surface, as more than a few of the leading participants refer to… and it seems that a lot of people are learning from the contest. In turn… it seems like a great wikinomics story, non? It’s hard to find the downside, unless one pursued the argument that some publicly subsidized academic institutions are paying people to work on NetFlix’s problem… but I don’t think it goes too far.  Particularly if you focus on the graduate students – many would be doing this kind of thing whether there was a contest or not, so a million bucks on the horizon sure doesn’t hurt.

If I was to offer one criticism, it would be in relation to the reward distribution. The contest, as it was set-up, really encourages secrecy and “proprietary” development. While a collaborative spirit has emerged and people are finding the value of teaming up, I wonder what would have happened if the incentive distribution structure implicitly encouraged collaboration? One idea would be splitting out half the prize money to (say) be shared amongst the community based on how much each contributed  and when (the earlier the better)- building on the Cambrian House/ Kluster model.



Comments are closed.

Now available in paperback!
Don Tapscott and Anthony D. William's latest collaboration, Macrowikinomics: New Solutions for a Connected Planet. Learn more.

Business - Oct 5, 2010 12:00 - 0 Comments

DRM and us

More In Business


Entertainment - Aug 3, 2010 13:14 - 2 Comments

Want to see the future? Look to the games

More In Entertainment


Society - Aug 6, 2010 8:19 - 4 Comments

The Empire strikes a light

More In Society