tea.mathoverflow.net - Discussion Feed ("What’s wrong with this proof?")

Greg Kuperberg comments on ""What’s wrong with this proof?"" (8235)

2010-08-10T09:34:57-07:00

I think that it's fine to ask once for references that discuss an attempted proof of a big conjecture. There have now been several questions posted on this theme, generally in a dubious form such as "What's wrong with this proof". But asking to a pointer to the right blog or wiki to see what's going on is fine, in my opinion.

Kaveh comments on ""What’s wrong with this proof?"" (8226)

2010-08-10T01:13:50-07:00

"Ramping up the hype" was not my intension, but I drop my position as there is now a polymath like attempt to discuss the paper.

Now any suggestion about what to do with the question I have posted? I would prefer to modify it to make it an independent question mainly about "how can an approach based on descriptive complexity avoid being a natural proof in the sense of Raborov-Rudich?" and "are there any previous results in descriptive complexity that avoid being Razborov-Rudich natural proofs barrier?" and remove the reference to Vinay's work from the title. I hope that this will save the question and remove the negative feeling and objections against it. Any opinions?

Pete L. Clark comments on ""What’s wrong with this proof?"" (8221)

2010-08-09T12:47:15-07:00

@Felipe: The question is now deleted.

voloch comments on ""What’s wrong with this proof?"" (8219)

2010-08-09T12:22:11-07:00

@Pete I just clicked on "delete". I am not sure whether it's done anything.

Pete L. Clark comments on ""What’s wrong with this proof?"" (8217)

2010-08-09T11:57:33-07:00

@Felipe Voloch: I think I agree with you that my question on the Datta paper and your answer are both somewhat regrettable. I have just unaccepted your answer. I believe you have to delete your answer (since it has been upvoted) before I can delete the question. I will do so if you do.

Scott Carnahan comments on ""What’s wrong with this proof?"" (8213)

2010-08-09T09:49:20-07:00

I am inclined to close the newer question, because I think blogs are a better format than StackExchange for asking about and disseminating breaking news and rumors. If you're wondering what Deolalikar's argument has to do with natural proofs, you'll probably see an explanation first on a blog, anyway. As far as I can tell, the only remaining purpose of the current question is to ramp up the hype.

Kaveh comments on ""What’s wrong with this proof?"" (8212)

2010-08-09T09:30:29-07:00

voloch: "Stop bumping old complexity theory questions. I think, despite you appearing to want to follow MO rules, you are really just trying to stir the pot"

Sorry if I have done anything wrong, but that was not intentional. I am not trying to stir the pot. I have only corrected some grammatical mistakes in my comment and added the cs.cc.complexity-theory tag to a few recent complexity questions, if this is what you mean by "bumping old complexity theory questions", I will stop doing this right away, though I don't understand what it has to do with this question. They were complexity questions lacking the tag, and they were on the firs page already. Have I done anything else? Here are my recent actions on MO: https://mathoverflow.net/users/7507?tab=recent.

voloch: "People should make a serious effort at reading the paper before asking such questions and they should refrain from underhanded methods to get attention to the issue."

As I said in my answer to Vicktor's comment on my question, I will try to follow this advice next time I ask a question (though it does not seem to be a MO rule or norm for asking questions), but it has nothing to do with the question itself. And could you please stop attacking me? What are "the underhanded methods to get attention to the issue" I am using? I have just asked a question, and I have come here to check if that will be an appropriate question before posting it. I really don't understand why you are reacting so negatively to this question.

volovh: "The natural places for this discussion are the TCS blogs, e.g. Lipton's or Aaronson's. I am sure you will have better luck over there. "

This is not a discussion, and I think it satisfies the requirements to be a an MO question, and it should be allowed to be on MO even if you think there are other natural places for them. The only serious and specific objection I have seen so far is that this is about an unpublished work, and the norm is to wait till it is published in a journal and it is disrespectful to discuss unpublished work of people before its publication. I personally would prefer to have opinion of some TCS researchers on this and see if they think it is appropriate or not, because it seems to me that the norm in TCS is a little different.

@Oliver: Thank you for the explanation. I understand your points. But the discussion about this work is already going on in public blogs which allow anonymous comments. And although I agree with your concerns, this has not happened yet, so I will appreciate if people wait a little longer to see if this gets out of control. I expect that no one will answer the question for a few days during which people working in complexity will read the paper and after that I might get a reasonable answer from some expert in the area. This is a very specific question and I don't think it will turn into a discussion about the work being correct or not.

voloch comments on ""What’s wrong with this proof?"" (8211)

2010-08-09T08:19:40-07:00

@Kaveh Stop bumping old complexity theory questions. I think, despite you appearing to want to follow MO rules, you are really just trying to stir the pot. I haven't yet voted to close your question about "natural proof", but I'm thinking I might. I have downvoted it. I predict it will not get any meaningful answers to it, unless there is something basic which is wrong in the paper. People having a serious crack at it will take a while to have an opinion. BTW, I have no opinion on the proof itself.

A while back, I answered a question of Pete Clark's about a proof of RH. I found an error in five minutes and that it why I answered. In retrospect, I think it was a mistake to ask and answer that particular question.

My position on the general issue is a bit softer than Olivier's and maybe there are specific questions on such proofs that are suitable for MO. People should make a serious effort at reading the paper before asking such questions and they should refrain from underhanded methods to get attention to the issue. I am sure many people are discussing this. The natural places for this discussion are the TCS blogs, e.g. Lipton's or Aaronson's. I am sure you will have better luck over there.

Bill Dubuque comments on ""What’s wrong with this proof?"" (8210)

2010-08-09T07:53:57-07:00

@VP I don't find your argument persuasive. There are plently of unpublished results that are discussed on MO and many other mathematical forums. Some may never be published - for a variety of reasons. If a question meets the rules then it is a legitimate question, regardless of its source. The rules should not include censorship based upon source. In the case at hand this point is probably moot since the question probably fails to meet other criteria without revision.

SimonPL comments on ""What’s wrong with this proof?"" (8209)

2010-08-09T06:01:22-07:00

I think that algori's reformulation "What are, in your opinion, potential difficulties with the strategy proposed in this paper?" is a good way to transfer knowledge between people who want to read and evaluate the article.

I would agree with Olivier's position if this were an ordinary research announcement, but this particular preprint is going to have to withstand a lot of scrutiny, and it feels reasonable to give people the opportunity to dissect it.

On the other hand, the discussion will probably get out of control and probably should be hosted by a welcoming blog.

Olivier comments on ""What’s wrong with this proof?"" (8208)

2010-08-09T05:59:43-07:00

@Kaveh Forgive me, I guess this is because I am a junior researcher, but I don't understand this, can someone be more specific?
@algori I'm not sure I understand the proof of lemma 15.3, specifically the statement such and such appears a bit strange. Could someone please help me with that?

Here is my impression: barring a clear mistake, commenting on a research-level work above the level of simple gossip is something very hard. Apart from experts, and even more specifically apart from experts having read the preprint in question, not many people will be able to judge if the answer given on MO is really correct or not (even lemma 15.3 could use non-standard notations and be correct after all, even though in isolation it seems clearly wrong). So answers could linger on MO asserting that some parts of the proof are correct, other sketchy, other flat out wrong and precious few people will be able to decide for sure if this is true or not. I think discussing the work of someone else in a public forum which accepts anonymous participation without making sure that strong guidelines are in force (and I can't see how these guidelines could be implemented in MO) is potentially disrespectful to the people having produced the work in question, and I can't imagine the net gain produced by this kind of discussion outweighing this potential disrespect. Of course, private discussions are OK and giving your opinion on your own website is OK.

By the way, and in fairness to Kaveh, this comment implies that yes, I have down voted the question because of what I understand is a social norm, albeit not one specific to mathematics. On the purely mathematical level, and for instance if the paper had been accepted, I would be very glad to know the answers to Kaveh's question.

algori comments on ""What’s wrong with this proof?"" (8207)

2010-08-09T05:21:46-07:00

VP, Kevin -- I think there may be a reasonable question along these lines. Example: "What are, in your opinion, potential difficulties with the strategy proposed in this paper?" Olivier -- one can always phrase it differently. Example: "I'm not sure I understand the proof of lemma 15.3, specifically the statement such and such appears a bit strange. Could someone please help me with that?" In this particular case, the cats are already out of the bag: the paper is available, if not on the arxiv. When people are discussing someone's private communication without prior accord, this is unacceptable. Politely discussing something that's publicly available is a different matter. If the proof works, it works, if it doesn't, it doesn't, and there is not much anyone can do about it.

Kaveh comments on ""What’s wrong with this proof?"" (8206)

2010-08-09T05:17:45-07:00

VP (Vicktor Protsak?): "In particular, reviews are not openly disseminated until the paper has been published in a journal. There are very, very good reasons not to deviate from these norms." Oliver: "commenting on an unpublished work is a delicate matter, raising hard personal and technical issues."

Forgive me, I guess this is because I am a junior researcher, but I don't understand this, can someone be more specific?

Other people have already started commenting on this work. I guess this difference in opinion might be due to a difference between TCS (and CS) culture and the norms in other areas in mathematics. Please remember that in TCS publishing in journals is not the norm, many papers never appear in journal versions (I can cite a number of discussions by TCS people on this if needed). IMHO, it might be nice if we have more opinions on this from people working in TCS to see if this kind of questions are against the accepted norms in TCS.

ps: if the reason that you don't like a question is this norm in mathematics, please state so in your comments. It is not nice nor helpful to state other reasons when the real reason you are voting a question down or objecting to it is this.

Olivier comments on ""What’s wrong with this proof?"" (8205)

2010-08-09T04:47:57-07:00

@Suresh and more generally

"To follow up on Kaveh, I have been reading the paper and have an even more specific question relating to background knowledge about phase transitions for random k-SAT. Would that be admissible ?"

I deeply agree with VP: commenting on an unpublished work is a delicate matter, raising hard personal and technical issues. Thus, I would suggest that all questions of the form "lemma 15.3 of this paper seems wrong for this and this reason. Am I right?" should be closed, even if there are perfectly precise and technical. If the question can be formulated with no reference to the paper and if someone who has never heard of the paper could understand it, then of course things are different.

Suresh Venkat comments on ""What’s wrong with this proof?"" (8204)

2010-08-09T00:41:29-07:00

I was referring to Kaveh's question, not the original post ....

Kevin Buzzard comments on ""What’s wrong with this proof?"" (8203)

2010-08-09T00:36:33-07:00

@Suresh: "...an even more specific question". The original question could barely be less specific! "Is there a problem in these 100 pages of mathematics?". I agree with all the others in that MO is a terrible place to try and answer the original question. But if you have a specific *mathematics* question about something, and if that question has a mathematical answer (i.e. one that involves actual mathematics, not one that is just vaguely *about* mathematics), then sure, go ahead and ask it.

Suresh Venkat comments on ""What’s wrong with this proof?"" (8201)

2010-08-09T00:10:32-07:00

To follow up on Kaveh, I have been reading the paper and have an even more specific question relating to background knowledge about phase transitions for random k-SAT. Would that be admissible ?

Kaveh comments on ""What’s wrong with this proof?"" (8200)

2010-08-08T22:15:16-07:00

Is it OK if I ask a more specific question like:

does his approach avoid being a natural proof?

By the way, here is Dick Lipton's related post: http://rjlipton.wordpress.com/2010/08/08/a-proof-that-p-is-not-equal-to-np/

And another closed MO question: https://mathoverflow.net/questions/34947/when-would-you-read-a-paper-claiming-to-have-settled-a-long-open-problem-like-p/34957

Andy Putman comments on ""What’s wrong with this proof?"" (8194)

2010-08-08T16:10:14-07:00

I agree strongly with the points VP made and just cast the deciding vote to close.

Qiaochu Yuan comments on ""What’s wrong with this proof?"" (8193)

2010-08-08T16:02:05-07:00

@VP: Good points. I defer to your expertise.

VP comments on ""What’s wrong with this proof?"" (8192)

2010-08-08T15:58:17-07:00

Qiaochu, are you suggesting that MO be used as a repository of expert opinions on unpublished papers? Even if "there are experts who might be able to evaluate the content", there are long-standing procedures in mathematics for doing this sort of work. In particular, reviews are not openly disseminated until the paper has been published in a journal. There are very, very good reasons not to deviate from these norms. In addition, some features of MO, such as open participation and anonymous voting, make it even less suitable for publicizing opinions than blogs and personal websites.

Gjergji comments on ""What’s wrong with this proof?"" (8191)

2010-08-08T15:54:35-07:00

I really want to know the answer to the question, however I don't see what purpose it serves on MO at this point. The specialists are probably already aware of this paper and so it seems like a good idea to close the question as it stands and wait to reopen till a more solid opinion is formed. The question currently needs some editing anyways.

Qiaochu Yuan comments on ""What’s wrong with this proof?"" (8190)

2010-08-08T15:25:50-07:00

I'm torn. On the one hand, there are several experts here who might be capable of evaluating the content of this paper, and MO would be a valuable place to store those evaluations. On the other hand, this feels a little gossipy. I guess I'll wait and see what everyone else thinks.

theojf comments on ""What’s wrong with this proof?"" (8189)

2010-08-08T15:21:10-07:00

As written, http://mathoverflow.net/questions/34953/whats-wrong-with-this-proof is an awful question. The title has almost no information, and sounds crankier than it is, and the question is not a deep one. I don't see immediately how to salvage the question.

But perhaps it should be salvaged. I have no voted to close (yet), but I have added a comment linking to this thread.