Bear in mind that in many journals you'd have to go through that whole process for each individual paper
OK, but if it is already happening under the current system then we have not really taken any steps backward. If that is not happening under the current system, why would it happen under the system I described?
a random subset could quite well include people who have no qualms with actively screwing over an author they hate/compete with.
I know for a fact that happens under the current system; it happened to one of the people in my own research group. It is unfortunate, by as was noted elsewhere there are limits to how anonymous authors and reviewers can actually be, especially in very specialized fields where everyone knows what everyone else is working on.
I think points (a) and (b) cannot really be solved by any technical means; if universities are not interested in facilitating peer review or editing and would rather just continue to pay publishing companies to do so, then even a perfect technical solution is irrelevant.
That being said, point (c) can be addressed by having many institutions collaborate on managing journals. I do not think that this is inherently problematic, and if the process is completely transparent then conflicts of interest could be immediately pointed out. It is not perfect, but neither is the current system; I think getting rid of publishing companies would ultimately be beneficial, and that it would not be too hard to mitigate/manage the problems of universities running the peer review process.
As for accepting papers from people who are not affiliated, I am not understanding your point -- are you claiming that if universities were running the peer review process, unaffiliated researchers would be unable to publish their work? If the submissions were anonymous to begin with, how would that even work? I envision a system where papers are submitted through some sort of a mix network, reviewed and given a group signature, and then published. The author names in the paper could be (cryptographically) committed, then opened once the authors are notified of their paper passing peer review (i.e. when the authors receive the signature, they open their commitment; the final paper would be the submitted paper, the signature, and the key to open the authors' names and contact information). We have the technology to do all of the above right now; it is really a matter of addressing your first two points, which I will admit may be fatal to this idea.
In the consortia model, if "University of California" (10,000 professors?) signs a review, there's a good chance someone knowledgeable in their field could narrow it down to 2 or 3 professors before reading it, and know exactly who wrote it after reading it.
Sure, but this is not something that will be true regardless of what system you use to manage peer review. My only point is that we can and should take publishing companies out of the loop -- they serve no purpose that cannot be served better / at lower cost using the Internet. The only requirement is that we do not weaken the security that publishing companies provide as a service right now -- in technical terms, I should be able to simulate a publishing company facilitating the peer review process, and such a simulation should not be distinguishable from a publishing company actually facilitating the process.
Also, authors are often asked to suggest some of the reviewers, and are allowed to preclude personal enemies/competitors.
The authors should not, however, know which, if any, of the suggested reviewers were actually selected.
AI research is haunted...
on
Where's HAL 9000?
·
· Score: 4, Interesting
Too many decades of lofty promises that never materialized has turned "AI research" into a dirty word...
Dropbox is similarly secure if you store an encrypted container.
This is not officially supported by Dropbox, however, and is very much ad-hoc. It also requires the user to take the time to configure such a system, unless your IT staff is going to do it for you, and even then you have the problem of users trying to use Dropbox for things that IT did not set up for them. Anything that adds hurdles to people doing their work is a potential security problem; it is easier to simply ban dropbox entirely than to have a policy that requires people to try to do things manually.
Which is a technique that is used by China as part of their censorship apparatus. Delisting search results is still an attack on free speech, even if the origin website remains accessible; if you are wondering just how important search engine results are, consider the effort spent on "search engine optimization" as well as the money Google makes displaying sponsored search results.
academia operates on reputation, so a paper in a more rigorous journal does and should carry more weight. This is a good thing because it helps separate good work from the junk
Except that I have seen good work published in less "rigorous" journals (well, in CS it is more conferences than journals, but the effect is the same: I have seen good work presented at lower-tier conferences, and I have even seen people cite groundbreaking papers and get published in top-tier conferences for incremental improvements, when the groundbreaking paper itself was rejected from the top tier conferences). The name of the journal that a paper is published in is only loosely related to the quality of the paper. Articles should be judged on the quality of the work, and reputations should be built by publishing quality results, not by publishing in particular journals.
You say things like "papers can be..." but I'm saying that online access is worthless unless there is a certified transparent peer review process. I await your suggestions...
Well, we are having this conversation in two different threads, so I'll just reiterate this point: we can use group signatures to facilitate peer review. Researchers / institutions can coordinate to bring together a couple dozen reviewers, and then some random subset can review a particular paper; if the paper passes review, that subset applies a group signature for the group of reviewers and concatenates it to the paper, which is then published, or else sends anonymous feedback to the authors of the paper (which is what happens in the current system). When someone reads a paper, they can see that some subset of the group of reviewers agreed that the paper should pass the review process, but learns nothing about that subset. The only thing that needs to be done is for the researchers to be organized, but that is something that can be done cooperatively and which does not require a publishing company to facilitate.
That is what DOI is for, and there is no reason that big research institutions could not store archives of published work -- which they already do by maintaining bookshelf upon bookshelf of bound journals.
Well, I think the solution lies in cryptography (disclaimer: I am a grad student doing research in cryptography). You need a system where researchers in a field could apply digital signatures to papers, but with a twist: the reviewers should remain anonymous after applying those signatures. This is not an impossible task; it is called a "group signature." The idea is that universities/researchers would cooperate to bring peer reviewers together, and those reviewers would be given group signature keys that they would apply to the papers they review. A person reading a paper could verify the signatures, which would tell them which consortium of universities/researchers organized the review process for that paper.
Like journals, the groups of reviewers could be organized on a per-month basis, and the names the whole group would be published -- with only a fraction actually reviewing any particular paper. It is not a complete break from journals as a system, it is just a way to use computers and the Internet to publish instead of relying on the old publishing companies; the way researchers communicate with each other has changed, and publishing articles should change too.
Not necessarily a step backwards; it could be a step towards ending the anachronism that is journal publishing. Really, what are journals doing for us these days, that cannot be done online?
Researchers can be organized over the Internet to participate in peer review; this is already done voluntarily in many cases under the current system.
Editing can be coordinated online as well, and is likewise often done by unpaid volunteers.
Papers can be distributed to researchers over the Internet instead of being bound and printed.
So really, the only thing that journals have left at this point is their names -- a paper in a "top journal" looks good on a CV, regardless of whether or not the paper is really groundbreaking. Is that really something that justifies the continued existence of journals? I think not...
Now, let's get other big institutions on board with this, and then let's turn to the problem of journals. We really do not need journals anymore; their primary function is to distribute papers to other researchers, which can be done online, and peer review and editing can be done by professors at universities (and this is frequently the case anyway -- often unpaid). The Internet connects researchers to each other, so why are we not using it to accomplish these goals?
I'll no longer have to subsidize all the folks who watch networks like ESPN that I don't care about.
Actually, ESPN forces cable companies into an all-or-nothing situation: either all their customers pay for ESPN, or nobody gets to watch it. They are doing the same thing to ISPs, so it is not as if moving everyone onto the web will somehow help us escape this practice.
I fail to see the downside here...we'll simply be paying those fees directly to the entertainment companies
No thanks -- those are the same companies that want to kill the Internet. What we really need is to kill off the entire distribution chain, and build a new, more open system.
While it's possible to read these as humans, I think a more reasonable reading is that they're computers.
I do not agree with that; that is a common use case, but patents do not apply only to common use cases.
That one's even easier. Claim 1 recites a cryptosystem, which is illustrated in FIG. 2 including an ALU and RAM.
Fair enough, although this is still very vague and non-specific. It is basically saying that the patent covers any electronic computer that runs the algorithm -- which includes computers that have not even been invented yet. If you had no familiarity with computers, and I asked you to describe the machine that this patent covers, I doubt that you would be able to give an answer that even remotely resembles the sorts of machines that ECC is commonly use on.
Claim 1: "1. A method of generating in a digital data processor..."
Like I said, if you did not know what a computer is and you were asked to describe the machine that this patent covers, you would be unable to give a description that would be anything close to the machines that ECC is used on.
There you go. In all cases, pure math - or performing the steps in your head or on a pad of paper - would not be covered by the patent claims. Therefore, they can't be claiming pure math.
Except that they contain no specific descriptions of any machine, only vague and abstract references to machines, and the rest of the claims concern abstract mathematical operations. You might have a point if, say, any one of the patents mentioned even one real-world computer architecture in its claims. They do not go that far. As I originally said, these are nothing more than formal statements about computers; the core of the patent, and the patent itself, cover pure math.
Even if these patents named a specific computer that the algorithm is run on, the patents would still cover the mathematics of the algorithm and the machine would be irrelevant. These are not patents on a sequence of steps that a machine performs; if the same algorithms described in the patents were compiled two different ways i.e. two different sequences of instructions being carried out by the target computer, both programs would violate the patent. Some of those patents mention using the quadratic formula; yet there are infinitely many sequences of operations that a computer could perform to compute that formula, and so these patents do not cover any particular sequence. It is not possible to avoid violating these patents by, say, changing your compiler flags to produce different assembly language sequences. There is no specific sequence of operations covered, only the abstract math that certain sequences implement, and that is on any computer architecture.
We are also completely ignoring the original point about industrial processes being processes that concern the transformation of some physical material. At no point do any of these patents mention physical materials. These patents do not cover industrial processes by any stretch of the imagination; they cover software being run on any computer, not necessarily electronic computers, and including computers that have not been invented yet. Again, these patents cover the abstract algorithm; the mention of computer hardware is only formal.
You are missing the point about throwing grenades into the homes of people who committed no crime, then busting down their door and shooting them. This happens all over the country:
The SWAT teams and drug squads who kill people generally receive no punishment for their actions, even when they kill innocent people. These are not actions that benefit society -- these are people who committed no crime and posed no danger, who were killed by paramilitary police squads.
Can you even see what machine this patent refers to? The closest thing to that is in claim 122,
a data communication system
Which may not even refer to a machine, since we can communicate by shouting at each other across a room, by writing numbers of sheets of paper, etc. This is not even a formal reference to using a computer; it is just a vague reference to the concept of communicating electronically. Otherwise, these claims all cover pure math.
As if this is news? For decades, the feds have been busting into innocent people's homes and killing them and their dogs, without being arrested or imprisoned for it. Now, if I were to throw a grenade into someone's house, rush in with an assault rifle and kill them, what do you suppose would happen to me?
Or any of the hundreds of other patents on mathematics? What specific machine or material does a patent on finite field representations cover? Keep in mind that Certicom claims that its patents cover all computer architectures equally.
These are not industrial processes, these are not specific machines, these are patents on pure math with a formal statement about running the software on some computer.
Sure, and that was the original logic behind the first software patent. The difference is that today's software patents are not on industrial processes, they are on pure mathematics, which is not supposed to be patentable.
Except that the process clause is meant to cover industrial processes that are not strictly machines, but which are physical inventions nevertheless (at least in the sense that they transform physical material from one form to another).
How is the concept of computing something substantially different from an algorithm that computes something? Patents are supposed to be on physical inventions, not abstract ideas. The formal, "When run on a computer," clause does not mean a math^H^H^H^Hsoftware patent is somehow not a patent on an abstract idea.
Slashdot Mirror
Bear in mind that in many journals you'd have to go through that whole process for each individual paper
OK, but if it is already happening under the current system then we have not really taken any steps backward. If that is not happening under the current system, why would it happen under the system I described?
a random subset could quite well include people who have no qualms with actively screwing over an author they hate/compete with.
I know for a fact that happens under the current system; it happened to one of the people in my own research group. It is unfortunate, by as was noted elsewhere there are limits to how anonymous authors and reviewers can actually be, especially in very specialized fields where everyone knows what everyone else is working on.
I think points (a) and (b) cannot really be solved by any technical means; if universities are not interested in facilitating peer review or editing and would rather just continue to pay publishing companies to do so, then even a perfect technical solution is irrelevant.
That being said, point (c) can be addressed by having many institutions collaborate on managing journals. I do not think that this is inherently problematic, and if the process is completely transparent then conflicts of interest could be immediately pointed out. It is not perfect, but neither is the current system; I think getting rid of publishing companies would ultimately be beneficial, and that it would not be too hard to mitigate/manage the problems of universities running the peer review process.
As for accepting papers from people who are not affiliated, I am not understanding your point -- are you claiming that if universities were running the peer review process, unaffiliated researchers would be unable to publish their work? If the submissions were anonymous to begin with, how would that even work? I envision a system where papers are submitted through some sort of a mix network, reviewed and given a group signature, and then published. The author names in the paper could be (cryptographically) committed, then opened once the authors are notified of their paper passing peer review (i.e. when the authors receive the signature, they open their commitment; the final paper would be the submitted paper, the signature, and the key to open the authors' names and contact information). We have the technology to do all of the above right now; it is really a matter of addressing your first two points, which I will admit may be fatal to this idea.
In the consortia model, if "University of California" (10,000 professors?) signs a review, there's a good chance someone knowledgeable in their field could narrow it down to 2 or 3 professors before reading it, and know exactly who wrote it after reading it.
Sure, but this is not something that will be true regardless of what system you use to manage peer review. My only point is that we can and should take publishing companies out of the loop -- they serve no purpose that cannot be served better / at lower cost using the Internet. The only requirement is that we do not weaken the security that publishing companies provide as a service right now -- in technical terms, I should be able to simulate a publishing company facilitating the peer review process, and such a simulation should not be distinguishable from a publishing company actually facilitating the process.
Also, authors are often asked to suggest some of the reviewers, and are allowed to preclude personal enemies/competitors.
The authors should not, however, know which, if any, of the suggested reviewers were actually selected.
Too many decades of lofty promises that never materialized has turned "AI research" into a dirty word...
Dropbox is similarly secure if you store an encrypted container.
This is not officially supported by Dropbox, however, and is very much ad-hoc. It also requires the user to take the time to configure such a system, unless your IT staff is going to do it for you, and even then you have the problem of users trying to use Dropbox for things that IT did not set up for them. Anything that adds hurdles to people doing their work is a potential security problem; it is easier to simply ban dropbox entirely than to have a policy that requires people to try to do things manually.
Which is a technique that is used by China as part of their censorship apparatus. Delisting search results is still an attack on free speech, even if the origin website remains accessible; if you are wondering just how important search engine results are, consider the effort spent on "search engine optimization" as well as the money Google makes displaying sponsored search results.
academia operates on reputation, so a paper in a more rigorous journal does and should carry more weight. This is a good thing because it helps separate good work from the junk
Except that I have seen good work published in less "rigorous" journals (well, in CS it is more conferences than journals, but the effect is the same: I have seen good work presented at lower-tier conferences, and I have even seen people cite groundbreaking papers and get published in top-tier conferences for incremental improvements, when the groundbreaking paper itself was rejected from the top tier conferences). The name of the journal that a paper is published in is only loosely related to the quality of the paper. Articles should be judged on the quality of the work, and reputations should be built by publishing quality results, not by publishing in particular journals.
You say things like "papers can be..." but I'm saying that online access is worthless unless there is a certified transparent peer review process. I await your suggestions...
Well, we are having this conversation in two different threads, so I'll just reiterate this point: we can use group signatures to facilitate peer review. Researchers / institutions can coordinate to bring together a couple dozen reviewers, and then some random subset can review a particular paper; if the paper passes review, that subset applies a group signature for the group of reviewers and concatenates it to the paper, which is then published, or else sends anonymous feedback to the authors of the paper (which is what happens in the current system). When someone reads a paper, they can see that some subset of the group of reviewers agreed that the paper should pass the review process, but learns nothing about that subset. The only thing that needs to be done is for the researchers to be organized, but that is something that can be done cooperatively and which does not require a publishing company to facilitate.
That is what DOI is for, and there is no reason that big research institutions could not store archives of published work -- which they already do by maintaining bookshelf upon bookshelf of bound journals.
Well, I think the solution lies in cryptography (disclaimer: I am a grad student doing research in cryptography). You need a system where researchers in a field could apply digital signatures to papers, but with a twist: the reviewers should remain anonymous after applying those signatures. This is not an impossible task; it is called a "group signature." The idea is that universities/researchers would cooperate to bring peer reviewers together, and those reviewers would be given group signature keys that they would apply to the papers they review. A person reading a paper could verify the signatures, which would tell them which consortium of universities/researchers organized the review process for that paper.
Like journals, the groups of reviewers could be organized on a per-month basis, and the names the whole group would be published -- with only a fraction actually reviewing any particular paper. It is not a complete break from journals as a system, it is just a way to use computers and the Internet to publish instead of relying on the old publishing companies; the way researchers communicate with each other has changed, and publishing articles should change too.
So really, the only thing that journals have left at this point is their names -- a paper in a "top journal" looks good on a CV, regardless of whether or not the paper is really groundbreaking. Is that really something that justifies the continued existence of journals? I think not...
Now, let's get other big institutions on board with this, and then let's turn to the problem of journals. We really do not need journals anymore; their primary function is to distribute papers to other researchers, which can be done online, and peer review and editing can be done by professors at universities (and this is frequently the case anyway -- often unpaid). The Internet connects researchers to each other, so why are we not using it to accomplish these goals?
In any case, this is a good first step.
infringement obviously needs to be taken down quickly
Obviously for you, maybe. Copyright infringement is supposed to be decided by courts, which is not an 11 hour process.
I'll no longer have to subsidize all the folks who watch networks like ESPN that I don't care about.
Actually, ESPN forces cable companies into an all-or-nothing situation: either all their customers pay for ESPN, or nobody gets to watch it. They are doing the same thing to ISPs, so it is not as if moving everyone onto the web will somehow help us escape this practice.
I fail to see the downside here...we'll simply be paying those fees directly to the entertainment companies
No thanks -- those are the same companies that want to kill the Internet. What we really need is to kill off the entire distribution chain, and build a new, more open system.
Try searching for, "This document is confidential" on Google. You would be surprised by what sort of things turn up.
If they were "nuking" Al Qaeda websites, maybe they would not get in trouble.
While it's possible to read these as humans, I think a more reasonable reading is that they're computers.
I do not agree with that; that is a common use case, but patents do not apply only to common use cases.
That one's even easier. Claim 1 recites a cryptosystem, which is illustrated in FIG. 2 including an ALU and RAM.
Fair enough, although this is still very vague and non-specific. It is basically saying that the patent covers any electronic computer that runs the algorithm -- which includes computers that have not even been invented yet. If you had no familiarity with computers, and I asked you to describe the machine that this patent covers, I doubt that you would be able to give an answer that even remotely resembles the sorts of machines that ECC is commonly use on.
Claim 1: "1. A method of generating in a digital data processor..."
Like I said, if you did not know what a computer is and you were asked to describe the machine that this patent covers, you would be unable to give a description that would be anything close to the machines that ECC is used on.
There you go. In all cases, pure math - or performing the steps in your head or on a pad of paper - would not be covered by the patent claims. Therefore, they can't be claiming pure math.
Except that they contain no specific descriptions of any machine, only vague and abstract references to machines, and the rest of the claims concern abstract mathematical operations. You might have a point if, say, any one of the patents mentioned even one real-world computer architecture in its claims. They do not go that far. As I originally said, these are nothing more than formal statements about computers; the core of the patent, and the patent itself, cover pure math.
Even if these patents named a specific computer that the algorithm is run on, the patents would still cover the mathematics of the algorithm and the machine would be irrelevant. These are not patents on a sequence of steps that a machine performs; if the same algorithms described in the patents were compiled two different ways i.e. two different sequences of instructions being carried out by the target computer, both programs would violate the patent. Some of those patents mention using the quadratic formula; yet there are infinitely many sequences of operations that a computer could perform to compute that formula, and so these patents do not cover any particular sequence. It is not possible to avoid violating these patents by, say, changing your compiler flags to produce different assembly language sequences. There is no specific sequence of operations covered, only the abstract math that certain sequences implement, and that is on any computer architecture.
We are also completely ignoring the original point about industrial processes being processes that concern the transformation of some physical material. At no point do any of these patents mention physical materials. These patents do not cover industrial processes by any stretch of the imagination; they cover software being run on any computer, not necessarily electronic computers, and including computers that have not been invented yet. Again, these patents cover the abstract algorithm; the mention of computer hardware is only formal.
You are missing the point about throwing grenades into the homes of people who committed no crime, then busting down their door and shooting them. This happens all over the country:
http://www.drugwarrant.com/articles/drug-war-victim/
The SWAT teams and drug squads who kill people generally receive no punishment for their actions, even when they kill innocent people. These are not actions that benefit society -- these are people who committed no crime and posed no danger, who were killed by paramilitary police squads.
http://www.google.com/patents?vid=6563928
Can you even see what machine this patent refers to? The closest thing to that is in claim 122,
a data communication system
Which may not even refer to a machine, since we can communicate by shouting at each other across a room, by writing numbers of sheets of paper, etc. This is not even a formal reference to using a computer; it is just a vague reference to the concept of communicating electronically. Otherwise, these claims all cover pure math.
http://www.google.com/patents?vid=6782100
Tell me if you can find a claim that even vaguely refers to a machine, because I cannot.
http://www.google.com/patents?vid=5854759
Again, I am not seeing a claim that mentions a machine, but maybe you can point it out for me.
The ball is in your court; tell us how these patents are somehow not actually patents on pure math.
As if this is news? For decades, the feds have been busting into innocent people's homes and killing them and their dogs, without being arrested or imprisoned for it. Now, if I were to throw a grenade into someone's house, rush in with an assault rifle and kill them, what do you suppose would happen to me?
There aren't any "pure mathematics" patents.
What about these:
https://en.wikipedia.org/wiki/ECC_patents
Or these:
http://blogs.teamb.com/craigstuntz/2012/04/04/38707/
Or any of the hundreds of other patents on mathematics? What specific machine or material does a patent on finite field representations cover? Keep in mind that Certicom claims that its patents cover all computer architectures equally.
These are not industrial processes, these are not specific machines, these are patents on pure math with a formal statement about running the software on some computer.
Sure, and that was the original logic behind the first software patent. The difference is that today's software patents are not on industrial processes, they are on pure mathematics, which is not supposed to be patentable.
Except that the process clause is meant to cover industrial processes that are not strictly machines, but which are physical inventions nevertheless (at least in the sense that they transform physical material from one form to another).
How is the concept of computing something substantially different from an algorithm that computes something? Patents are supposed to be on physical inventions, not abstract ideas. The formal, "When run on a computer," clause does not mean a math^H^H^H^Hsoftware patent is somehow not a patent on an abstract idea.
Try as I might, I seem to be unable to get a patent on Euclid's algorithm.