Profile cover photo
Profile photo
Timothy Gowers
20,116 followers -
Mathematician
Mathematician

20,116 followers
About
Communities and Collections
View all
Posts

Post has attachment
A few weeks ago I posted about the Journal of Algebraic Combinatorics "flipping" -- that is, leaving its existing publisher (Springer) and setting up with a modified name and complying with Fair Open Access principles. They have recently put out a press release to give an update on how things are going. And the answer is that they are going swimmingly, with lots of good quality submissions and a great deal of support from mathematicians for the whole enterprise. The really important thing about this was that Springer had not behaved especially badly: they just did what one would expect of a company that belongs to a whole system that is bad, and that we should not support.

Post has attachment
*How do you value a pension fund?*

I don't begin to know the answer to this, or how to evaluate an answer somebody else has given. But if you belong to the Universities Superannuation Scheme in the UK, then it is an important question. There have been headlines recently suggesting that the USS's assets are a long way short of what is needed to meet their liabilities -- that is, the pensions of people who have paid into their scheme. They have already made the scheme less generous, including what I see as a breach of contract, even though formally it isn't one. (In brief, I paid into a final-salary scheme. They stopped the scheme, and the payments already made will lead to payouts based on one's salary at the time that the change was made. So I, and many like me, paid into the scheme on what turned out to be a false understanding. In the small print it said that they had the right to do this, but there was no suggestion when I started my career that this was a risk that I should take into account.) The latest evaluations could be an attempt to cut the payouts even further. But the estimates seem to vary wildly from one evaluation to another.

Recently there was a campaign to make all the data open so that academics could see the working that lay behind these evaluations -- after all, it was our money at stake. Amazingly, it has been successful. The link below is to the information, and I'm posting it in case anyone reading this has the expertise to make use of it.

Post has attachment
The difference between Norbert Blum and a crank

One obvious difference is that Blum has already established himself as an extremely reputable researcher: he has published serious results in the area and is very well acquainted with all the literature about techniques that cannot work to prove that P=/=NP, giving careful explanations of why they did not apply to his proof attempt. But even stronger evidence of non-crankdom is that when people pointed out that his proof couldn't work, instead of clinging to it, he retracted it, and now promises to write a detailed account of what is wrong. Similar classy behaviour was exhibited by Edward Nelson a few years ago when his attempted proof of the inconsistency of ZFC turned out to contain a somewhat subtle error.

(Hat tip to +Lance Fortnow.)

Post has attachment
I was recently contacted by Amir Asghari to say that he has set up a website in memory of Maryam Mirzahkani. I don't mean that it is a memorial website: rather, it has taken an important aspect of her life -- her interaction with other like-minded mathematicians at a young age -- and aims to facilitate such interactions for others, by the setting up of maths4maryams groups.

Post has shared content
Is the German DEAL negotiation with Elsevier good news or bad news?

If you follow these things, you'll know that negotiators from Germany have reached an impasse with Elsevier. That is because instead of accepting that any new deal should be a small modification of whatever they had already, they have decided to go all out for significant change.

Instead of buying subscriptions to specific journals, consortium members want to pay publishers an annual lump sum that covers publication costs of all papers whose first authors are at German institutions. Those papers would be freely available around the world; meanwhile, German institutions would receive access to all the publishers' online content.

Removing paywalls from a lot of papers has to be good news doesn't it? Well, perhaps, but in a subject like mathematics it doesn't make too much difference, since nearly all sensible mathematicians put their work on arXiv. So I want to know about the price that the Germans hope to pay. Here's what the article says about that.

The German delegation is guided by a simple formula: Take the number of papers with first authors at German institutes put out by a publisher and multiply that by a reasonable fee per paper. That's what Germany should pay the publisher—and the total is likely to be much lower than current spending on subscriptions.

So the idea is to pay a lot less and remove a lot of paywalls. That is certainly good news. But I still have two concerns. The first is this:

Now the parties need to agree on an acceptable fee per article. The lower bound is the average article processing fee charged by existing OA journals, somewhere around €1300. The German Research Foundation, the country's main science funding agency, has set an upper limit of €2000 per published article. "We won't be able to get to [€1300], but we need to start on a path of lower prices that gradually brings us closer,"

What I would prefer to see is a proper estimate of what it ought to cost to process a paper, and for that to be the basis of what is paid. This isn't a huge concern for me, since they estimate the cost for a subscription article is more like €4000-€5000, so this would still be a very large drop in what Elsevier receives, and getting from the current system to an ideal system in one go is an unrealistic aim.

My main concern is the idea of paying an annual lump sum to a publisher that then makes it free for academics in your country to publish with them. To see what's wrong with that, imagine that a new super-efficient publisher comes along that can process articles for €500. Universities will be unwilling to pay this when their academics can publish with Elsevier for no charge. (Of course, there is a charge, but it has been paid as a tiny part of a huge lump sum.) So a big deal like this has the effect of perpetuating the current dominance of the big commercial publishers.

This second concern is not a concern if the proposed deal really does multiply the number of papers by a charge per paper, rather than, say, estimating the number of papers that will be published and paying a lump sum based on that estimate. In other words, I see the following two features of a deal as critically important, to avoid serious market distortion. The second implies the first.

(i) The marginal cost to the German academic community of each paper published by a German researcher should be roughly equal to the average cost. (In particular, it definitely shouldn't be zero.)

(ii) The marginal cost to each individual institution of a paper published by a researcher in that institution should be roughly equal to the average cost per paper of the whole deal.

The importance of the second principle is that if the cost of each paper is spread out over all the institutions, then that again removes any disincentive for individual institutions to publish with Elsevier. Of course, the principle can easily be satisfied: just make sure that individual institutions contribute to the big deal in proportion to how much they publish with Elsevier.

I've slightly oversimplified here, because of the additional demand that Germans should all have access to Elsevier papers, which may be used in different amounts by different institutions. But it is still very important for institutions not to be shielded from the cost of the deal.
Günter Ziegler (Free University of Berlin) on negotiating with Elsevier:

It's like you're at a car dealer trying to buy a car, but the salesperson keeps trying to sell you a carriage....You tell him "I don't want a carriage, I want a car." And he says: "Well if you buy this carriage, we'll give you this horse for free."

http://www.sciencemag.org/news/2017/08/bold-open-access-push-germany-could-change-future-academic-publishing

#oa #openaccess #elsevier

Post has attachment
Not very classy from CUP

Apparently CUP agreed to requests from the Chinese government to block access in China to articles in their journal China Quarterly that talk about topics such as Tibet, Tiananmen Square, etc. Their argument is that they did this “to ensure that other academic and educational materials we publish remain available to researchers and educators”. It's conceivable that they are utilitarians and believe that this action is maximizing the good that they can do in China. But (i) utilitarianism is not appropriate here -- they should take a stand and let the censorship be done by the Chinese themselves rather than being complicit in it, and (ii) in any case it seems rather more likely that it what they are trying to maximize is not the good to humanity but their own profits.

Norbert Blum's claimed proof that P=/=NP appears to be fatally flawed.

For those who have not been following the discussions on this topic, here is a rough summary of the current state of affairs as I understand them (those last four words being an important qualification).

(i) Blum's strategy is to show that a certain technique, related to but not quite the same as, Razborov's method of approximations for obtaining lower bounds on the monotone circuit complexity of monotone functions (this means you're allowed AND gates and OR gates but not NOT gates) has the following interesting property: if you can use it to obtain a superpolynomial lower bound for the monotone circuit complexity of a monotone function, then that automatically implies a superpolynomial bound for the general circuit complexity.

(ii) The method related to Razborov's method can be used to show that the monotone circuit complexity of any monotone function that outputs 1 for all cliques of size k and 0 for all complete (k-1)-partite graphs is superpolynomial.

(iii) Tardos showed that there exists such a function with polynomial circuit complexity.

Facts (ii) and (iii) demonstrate that Blum's strategy cannot work.

This problem seems to have been pointed out first by Gustav Nordh in the discussion at cstheory stackexchange and noticed independently (and pretty well immediately after he looked at the paper) by Razborov himself. And it is not a wise person who bets against Razborov.

Post has attachment
Ever wondered what happens if you have a glorious sunset and a rainbow at the same time? Well here's the answer.

Post has attachment
The flipping of the Journal of Algebraic Combinatorics makes the news. (Hat tip to +Sabine Hossenfelder.)

One detail has now been clarified for me. The press release said that almost all the editors had agreed to flip. It turns out that no editors have expressed the intention to stay with the old journal: one is retiring and they have not managed to get a response from another.

Post has shared content
I'll be blogging about this very soon -- I am one of the "many people" referred to by Mark Wilson and am thrilled by this outcome -- but for now I just want to help get the news out there.
Another journal breaks free! The result of a lot of work by many people.
Wait while more posts are being loaded