Until a major chorus of experts chimes in with "Yup, this is curtains for this approach", I think it would be wise not to take too seriously any blog comments based on quick reactions to the paper.
...just as it would be unwise to take the paper itself too seriously until a chorus of experts tells us that it's looking pretty good.
Tao has little expertise in the area, and he isn't pretending to -- he's deferring to those who are more knowledgeable. He happens to frequent these particular blogs, but he isn't making any statements as to the correctness of the proof.
However, the linked comment is the first major statement from a well-cited researcher (aside from S Aaronson, who hasn't even read the paper yet)
One expert finds the flaw, calls it "unfixable" twice, and another seems to agree. This is good enough for social news -- which while not always reliable is often ahead of the mainstream press.
removing the paper doesn't necessarily mean that at all, he probably just wants finish editing it and getting peer feedback before he releases it officially
well he had actually posted an updated version yesterday, so if he had to update it he would put up a not. Regarding the legal question, i don't think it reflect poorly any way on HP Labs. Its not like he is making cranky claims like no one landed on moon etc. etc.
I thought his paper was showing P/=NP which is the already (more generally) accepted theory in computer fields. I'd understand it reflecting poorly if one of HP's researchers was claiming P=NP against the accepted wisdom. It would however reflect great if this guys paper is a proof of P/=NP or even if it's simply a step forward.
It would be extremely unwise of HP to shut him down - regardless of whether his proof works out in the end, this is a significant piece of work contributed for the further understanding of hard math problems - which is what HP ideally wants to represent. To say no to endeavours like this would be to say that they have no soul and no will for the enhancement of human knowledge.
He just posted a new version of the draft, and added this blurb to his home page:
"Vinay Deolalikar. P is not equal to NP. 6th August, 2010 (66 pages 10pt, 102 pages 12pt). Manuscript sent on 6th August to several leading researchers in various areas. You can find the current version here. Please note that the final version of the paper is under preparation and will be submitted to journal review."
So it seems this is the defender our star forward has to get past to hit the ultimate math-world goal. We can now root for our respective teams... Go defenders!
I'm rooting for defenders on the principle that the present proof doesn't seem illuminating. I'd prefer a proof that seemed to give insight into the nature of complexity itself. But it's still just a diversion...
If I can understand the complexity classes themselves as well as the consequences of P = or != NP (which I believe I can), then it seems like I should expect (or at least hope for) a similarly-comprehensible proof.
Contrast that with, say, special relativity. I can't even really understand the consequences of special relativity (length contraction, relativity of simultaneity, time dilation, etc.) beyond accepting that my intuitions are completely errant at relativistic speeds. Because of that, I don't really expect or hope for an intuitive proof or explanation of special relativity.
Granted. I do think problems in the domain of number theory are a bit different. I don't know of any quickly-explainable consequences to Fermat's Last Theorem, and in my first comment I was really thinking about consequences. With Fermat's Last Theorem, to my knowledge, the original problem is itself pretty arbitrary, so I'm not surprised to find its solution and proof to be complicated despite the simplicity of the problem statement.
Also, I'll preemptively mention that Gödel showed us that any formal problem is a problem in number theory. That said, a problem like P =? NP would almost surely be incomprehensible if expressed as an isomorphic problem in number theory. So my original point, clarified and rephrased, is that I would expect (or desire) the proof of a problem to be as intuitive as the statement of the problem combined with its consequences in the domain the problem is stated.
I don't know the history of mathematical theories, but in physics it has often been the case that something took years (or decades) for people to wrap their heads around as they were poking at the edges of understanding. But as they continued to do so, new ideas and concepts emerged which made a better mental framework for understanding them. Hence, for example, the work-energy theorem can now be proved by a high school senior, but in the day there was big debate over the factor of 1/2 in the term for kinetic energy.
So it may be with mathematics. Things like Fermat's Last Theorem and P != NP may be difficult to grasp simply within our current frameworks, but new ideas may give rise to new frameworks that makes them more digestible.
If you can just get it proved somehow then you can proceed with confidence towards these new ideas.
indeed. there are probably not many people on HN or people who are going to post comments on blogs that are qualified to actually weigh in on this type of thing.
...just as it would be unwise to take the paper itself too seriously until a chorus of experts tells us that it's looking pretty good.