I've asked quite simple and concrete questions regarding the issues MarkHays has asked, under this thread, and Jeremy Howard responded, saying that all should be OK. (that is I can keep working in my field, using methods I might have used in the contest)

I can see people are not satisfied with his response, or people do not think that his response matches with the written rules. I'll keep saying that introducing IP restrictions only for the winner or milestone winner entries would solve a lot of problems. If HPN claims rights to only winning entry, I'm sure people would feel much more comfortable.

 

That's right - the issues that have been raised have already been answered by me in this thread, pointing to the specific definitions in the rules and showing how they are used. Unfortunately, since then, some competitors have posted incorrect advice.

I recommend that competitors read the actual rules, and the explanations provided by HPN and Kaggle in this thread and in the FAQ.

Dear Jeremy:

I could not find specific answers to my questions in this long thread, and I have not received a reply to my emails to Kaggle -- which I posted at the beginning of the competition.  Could you please ask the management of Kaggle and HPN to provide clear answers to two questions, based on Section 21 of the Official Rules:

(1) Does HPN own an unlimited worldwide right to every "algorithm" submitted by every competitor -- whether they win a prize or not?

(2) Does this license allow HPN to sell every algorithm submitted by every competitor to any company worldwide, with no license fees to any competitor?

(3) According to the last sentence in Section 21, are all competitors prohibited from selling the algorithms they submit to the Competition -- whether they win a prize or not?

In short, it appears that HPN owns everything every competitor submits, and can sell our work worldwide -- with no compensation.  Is this correct?

It would be helpful if these simple questions are answered simply and directly.  As I noted in a previous post, I would have no problem with a no-cost license to HPN.  I am surprised, however, that this "X Prize for Healthcare" appears to claim ownership to every entry.  Imagine if Progressive Insurance told every company that entered their electric vehicle competition, "Thanks for entering.  Now give us your car, and the exclusive rights to sell it worldwide!"

Mark Hays 

Mark, you do provide HPN with a license, but it is not "unlimited". It is limited by the rules and the definitions. In particular, by the definition of "Entry" and "Prediction Algorithm". So for example, in answer to your question "are all competitors prohibited from selling the algorithms they submit to the Competition": you can not sell a total package that reads in the data, does the modelling, and exports your answer to a file to be submitted to Kaggle. Of course, it would make no sense for you to want to do this, so it's not likely to be a limitation in practice!

Dear Jeremy:

Thank you for the quick reply.  We need, however, direct answers to the simple questions in my previous post.  Again:

(1) Does HPN own a worldwide right to every "algorithm" submitted by every competitor -- whether they win a prize or not?

(2) Does this license allow HPN to sell every algorithm submitted by every competitor to any company worldwide, with no license fees to any competitor?

(3) According to the last sentence in Section 21, are all competitors prohibited from selling the algorithms they submit to the Competition -- whether they win a prize or not?

In summary, it appears that HPN owns everything every competitor submits, and can sell our work worldwide -- with no compensation.  Is this correct?

Mark Hays

Ah, excelent, a chance to get some of my question answered (All but the last of which have been asked previously, though not all in this thread.):

1) What is the line between the entity you claim exclusivity over and the component parts that you do not? It is hard to interpret your previous coments to mean anything other than even a slight change resulting in it being a different algorithm, by your interpretation of the definitions.

2) Whist a restriction on publishing research that uses the data is fair, what about an algorithm (In the real CS sense, not the rules strange one.) being developed for the competition and then published without using the data? This matters a lot, especially once the competition has ended.

3) Does Kaggle/HPN claim any ownership over the algorithms that go into a final solution? i.e. if I develop some new regression/classification algorithm and use it as part of an entry am I then free to do what I like with it? - publish papers, sell it and do further research on it.

4) At what level do you represent HPN? Obviously you are a Kaggle employee, so can we trust that what you are saying is what HPN actually means?

 

(Sorry if anyone saw a partial version of this message- tablets are not good for composing long messages!)

Jeremy, I think you need to re-read your own rule set.  Here is Rule 21, in it's entirety.  Note the first line which says by registering for the competition...

 {Begin Quoted Rule}

By registering for the Competition, each Entrant (a) grants to Sponsor and its designees a worldwide, exclusive (except with respect to Entrant), sub-licensable (through multiple tiers), transferable, fully paid-up, royalty-free, perpetual, irrevocable right to use, not use, reproduce, distribute (through multiple tiers), create derivative works of, publicly perform, publicly display, digitally perform, make, have made, sell, offer for sale and import each Entry and the Prediction Algorithm used to produce the Entry (collectively, the "Licensed Materials"), in any media now known or hereafter developed, for any purpose whatsoever, commercial or otherwise, without further approval by or payment to Entrant (the "License") and (b) represents that he/she/it has the unrestricted (except for ‘Open Source’ restrictions in OSI-approved licenses listed at http://www.opensource.org/licenses/alphabetical or when Sponsor, in its sole discretion, expressly waives such requirement in writing) right to grant the License. Entrant understands and agrees that the License is exclusive except with respect to Entrant: Entrant may use the Licensed Materials solely for his/her/its own patient management and other internal business purposes but may not grant or otherwise transfer to any third party any rights to or interests in the Licensed Materials whatsoever. 

{End Quoted Rule}

I should note that this is not what the rule originally read when I first signed up.  It is explicitly bait-and-switch.  Simply by entering, regardless of whether they pay you one dime, each and every entry belongs to HPN.  In other words: "All you base belong to us..."

I should also not that, in the original rules, the rule was "non-exclusive", rather than exclusive, and the last line about basically reselling rights did not exist.  I have no intention of posting one character of code  unless I'm 99.9% successful.

Great reply, FineLine: 

There are very few limits on the license granted to HPN.  Based on Section 21, it appears that they have the right to sell every algorithm submitted by every competitor.  Moreover, competitors surrender the right to sell their own work!

I hope that Kaggle / HPN management will respond quickly and clearly -- and modify the rules.  Our organization is not-for-profit, with a mission to improve healthcare for the underserved.  We would like to join the competition, to help HPN and reduce readmissions nationwide.

Mark Hays 

Great reply, FineLine: 

There are very few limits on the license granted to HPN.  Based on Section 21, it appears that they have the right to sell every algorithm submitted by every competitor.  Moreover, competitors surrender the right to sell their own work!

I hope that Kaggle / HPN management will respond quickly and clearly -- and modify the rules.  Our organization is not-for-profit, with a mission to improve healthcare for the underserved.  We would like to join the competition, to help HPN and reduce readmissions nationwide.

Mark Hays 

Ask a local lawyer in your country about http://en.wikipedia.org/wiki/Standard_form_contract and http://en.wikipedia.org/wiki/Unconscionability Chances are, you did not enter a binding contract by clicking on some (repeatedly changing) web form to download the data, and HPN has no leverage to make you hand over any algorithm. Unless you actually end up qualifying for a prize. In which case they can withhold payment to try to make you comply, of course.

Daniel: Very little of this is actually about what happens at an actual court case - it is about uncertainty. Given so much uncertainty it is impossible to know how we can and can not act, both during and after the competition. Such uncertainty makes lawyers, and people in general, very conservative - right now I can not enter this competition - the rules can be interpreted such that I can enter (Well, maybe), but because there are many interpretations, many of which are extremely risky to me, I act in the interest of my own legal safety and stay clear. But to more directly answer you they state that legal action would be under California law, and therefore held in California - I could not afford a lawyer, and equally could not afford to not turn up as I need to travel to the US for work. Local laws may protect me from the direct consequences of legal action, but not the indirect, especially when dealing with an entity with orders of magnitude more money on its home turf.

Can I make a simple recommendation here - could someone from Kaggle (Or even better HPN) other than Jeremy read through this thread and post a response? I can't help but feel that Jeremy just doesn't get it - his failure to see the ambiguity and omissions in both the rules and his responses is startling to say the least, not to mention the contradictions between what he is saying and what the rules actually say. (I know the above will come across as being somewhat personal, but I honestly can't think of a better way of saying it. All I want is some certainty, so I can determine if I can safely enter or not - I really do not think this is asking for too much.)

Until you get to the point where you win and HPN requests the algorithm for your winning entry, you never submitted your algorithm. You only submitted entries. Those are mostly worthless, except for determining the winner.

For any entry, it's trivial to provide any number of algorithms that produce it. For example, you can submit an algorithm that prints random numbers, or one that simply prints the entry from literals contained in the algorithm.

It's impossible to prove that this isn't the algorithm that produced the entry, or that there is some other algorithm that produced the entry, that you're holding back.

If you would later sell a modified version of the original algorithm to a third party, it would also be impossible to prove that this algorithm was in fact the one that produced the winning entry.

Obviously you would need to strip any reference to HPN's data from the algorithm, but a third party would want the algorithm to work on THEIR data, not HPN's, anyway.

I don't know why Jeremy only hints at this, but that's what I read between his lines.

Thanks for the links Daniel. That clarified things a lot. These agreements are written by large corporations to cover their asses. Nobody can enforce each Entrant to hand over their algorithms. However, if they hand you the prize or you use the data willy-nilly, then this would give HPN leverage to sue you, if necessary.

This reminds me hilarious South Park season opener - HumancentiPad.

I don't see any reasons not to enter into this competition except for egoistic reasons. You can later justify that you are the smartest and you didn't win this competition because of the agreement. That being said, consult your tech transfer department if you are in academia and/or are having doubts about IP rights.

 

Dear Daniel:

Thank you for the thoughtful post.  Four points should be clarified, however:

(1) HPN owns your work when you register to enter the Competition:  According to Section 21 of the Rules, every competitor grants HPN (and any company they "designate") a full license to the prediction algorithm you use to create your entry.  This occurs at the moment you register for the Competition.  Here is a copy of the key section of Section 21, with highlights:

"By registering for the Competition, each Entrant (a) grants to Sponsor and its designees a worldwide, exclusive (except with respect to Entrant), sub-licensable (through multiple tiers), transferable, fully paid-up, royalty-free, perpetual, irrevocable right to use, not use, reproduce, distribute (through multiple tiers), create derivative works of, publicly perform, publicly display, digitally perform, make, have made, sell, offer for sale and import each Entry and the Prediction Algorithm used to produce the Entry (collectively, the "Licensed Materials")..."

In short, every competitor grants a full license to HPN when you initially register for the Competition.  This type of  "click through" agreement is enforceable in the United States.

(2) You could be forced to deliver the actual algorithm:  If HPN decides to enforce this agreement against you (particularly in the United States), their legal team could launch 'discovery' and force you to disclose your email, all communication and work products related to the Competition, etc.  Simply sending in a bogus formula / algorithm would provide no protection.

(3) They own your work -- and can demand a copy -- whether you win or not:  According to Section 21, you grant a full license to HPN when you register, whether you win a prize or not.  This is one of the most surprising sections of the agreement.  As I noted previously, I would have no problem granting a license to HPN if we won a prize; after all, they funded the Competition.  Section 21 goes much further, however.

(4) You give up the right to sell or distribute your work:  This is the second surprising part of Section 21.  Not only does HPN own all of your work, you can't sell it to any other company, period.  You can't even give it away for free, to promote improved healthcare!  Here is the key clause:

"Entrant may use the Licensed Materials solely for his/her/its own patient management and other internal business purposes but may not grant or otherwise transfer to any third party any rights to or interests in the Licensed Materials whatsoever."

Again, you give up your rights the minute you register for the Competition -- you don't have to send in an Entry or win a prize.

I sincerely hope that HPN management will consider revising these terms.  Our not-for-profit university, for example, has an advanced predictive solution, and we would love to enter it in the Competition.  We would also be happy to grant a full license to HPN if we won a prize. 

Most experienced researchers and predictive analysis experts who read Section 21 carefully will also be detered. 

I am going to contact Dr. Richard Merkin, President and CEO of HPN, to see if we can resolve these issues.  If I receive a response, I will let the Forum know.

Mark Hays

 

Mark Hays: I know you are primarily concerned with ownership but my primary concern is one of publication (With ownership a close second!). In principal if the rules are changed such that you own your algorithm (In the real CS sense - I couldn't care less about the problem specific code.), then publication, as long as you don't use their data, without their permission should be possible. 'Without their permission' is important, as I can not afford for them to say no, and at some point the competition ends and they are no longer available to be asked. If you could consider that when writing your email I would appreciate it - in principal it should be fixed anyway, but they might attempt to meet half way, and that issue may not then be resolved. (Its more of an ambiguity - the section on research is not clear, and then there are the terms about 'previously published'.)