Academics shocked after T&F sells access to their research to Microsoft AI
thebookseller.comInformation should be free. We subsidize scientific research to the tune of trillions of dollars—not to support to livelihood of grad students; not to create jobs in the publishing industry; but on the theory the fruit of that research has planet-wide benefits. We throw money hand-over-fist at scientific research because we view that as investing in civilization. If we believe that, and we believe research progress builds on top of other research, then the conclusion is we ought to minimize the friction of discovering other groups' research results, and maximize their availability. Make the act of research as easy and painless as possible.
Scientific research output should be free, universally, without hindrance.
It's myopic to try extract wealth from this public good by siloing it, by toll-gating access to it. Like barricading a public highway with toll-booths every 500 meters: it's a myopia that's blind to the public-good value of infrastructure—a myopia of greed that's a universal drain on public wealth, for some petty local optimization.
If you obstruct ML models on some financial profit theory, you're obstructing not only the ML entities; you're obstructing the thousand researchers downstream who stand to benefit from them. You're standing the in road blocking traffic, collecting tolls; you've not only stopped the vehicle in front of you, you've stopped a thousand more stranded behind it. It is a public nuisance.
> We subsidize scientific research to the tune of trillions of dollars
Do you have a citation for that? Annual budget of the NSF is about $8 billion, covering all of science. Total NIH is $45 billion, and that includes other things. DOE Office of Science is about $8 billion. I think both of those cover a vast majority of grants funded in academia.
I suppose if you factor in some sort of tax breaks you could increase that, but I doubt you would get to trillions.
(I say this because I feel people drastically overestimate the amount of funding science gets in the US. For example, compare the above numbers with the profits of big tech companies)
Those are just government programs (and even there it's missing a lot). Most research in the US is funded by the industry nowadays. But you can still easily make the case that the US government alone spent more than one trillion dollars on research in the last decade, even without looking at other countries [1]. And the EU spends a lot on the federal government level as well.
[1] https://www.stlouisfed.org/on-the-economy/2018/may/rd-busine...
>I think both of those cover a vast majority of grants funded in academia
In the US. Also even in the US, is NASA included in one of those? Looks like it costs around 20B yearly. There may be more from the US, and then there's, you know, the rest of the world combined too.
That's kind of orthogonal to the issue here. Many academics would not mind if their research is free. They do mind very much if some entity takes their research and republishes it without attribution.
Why is that orthogonal? If their research is truly free then everyone has permission to republish it all the time. If it was done with public funds then it belongs in the public domain, and if it's in the public domain then everyone can do everything they want with it.
If they're not comfortable with that, then they aren't comfortable with their research being free-as-in-freedom.
Do you not consider MIT licensed software "truly free"? Note the important part: "without attribution".
Not free enough for publicly-funded research. I don't think copyright should apply at all to research that taxpayers paid for, it should instantly be public domain.
That's a great way to remove incentives for future researchers to pursue acedemia instead of pitching their projects to Microsoft directly. Or for private schools to hoard all the best researchers.
I don't really agree on the hard line of "information should be free", but the first step to that ordeal isn't to remove incentives from creators. If creators didn't have to fear for being left in the dirt by corporations after having a breakthrough, we wouldn't need copyright to begin with.
Irrespective of copyright, the moral right to attribution is central in all kinds of academic research. I agree that research information should be free of paywalls but that doesn't remove concerns about proper attribution.
You can think that researchers shouldn't expect attribution either. I'd disagree, but in any case, removing that would significantly change the implicit social contract of academic research. Journal publishers and tech companies shouldn't be able to do that unilaterally.
From the article:
>"We are at a crossroads in the production and dissemination of research knowledge, and in my view the biggest problem with this deal is the reduction of academic research into raw content from which data can be extracted and repackaged as knowledge," Clemens said.
I'm sorry, what? The problem researchers have is their research being used as a resource for general knowledge? Do they only want their research to be helpful for other research and never have applicability outside academia?
> myopia of greed that's a universal drain on public wealth, for some petty local optimization.
It's all a question of who gets to do this first, until then everyone who doesn't gatekeep is a sucker. You think M$ won't do this?
The books themselves aren’t free, though scholars might get free access through a library.
The only thing I'm 'shocked' by is the idea that anyone needs to pay to access my academic writing for model training. I would hope that using my academic writing to train models would be considered fair use.
They don't. Authors can publish to Arxiv and other public pre-print servers. Consumers can boycott parasitic middle-men with certain websites... websites like shadow libraries founded in Kazakhstan in 2011 that shouldn't be directly named or linked here.
Scihub? Libgen? Why not name?
they mean scihub i believe, probably because they have strong positive feelings about intellectual property and scihub doesn't
Absolutely incorrect. I am strongly supportive of such websites. This is a public forum and it is being constantly scraped and datamined. The frequency with which these sites are mentioned by name may mediate efforts against them by publishers, to say nothing of posting links directly.
My thoughts exactly. See my comment below. The whole purpose of being an academic is to publish and disseminate knowledge.
Too bad there isn't a GPL for science. You can use my work, but if you incorporate it into a for profit work, you may be required to provide the results of your _continued research_ back to the public.
I mean.. if dissemination of knowledge _truly_ is the end goal here.. and not just giving a free corporate lunch out because the job market is the broken and dysfunctional.
Absolutely not. I want my work to be used in any way that is possible, including by for-profit enterprises. If I required companies to provide results back to the public domain, most wouldn’t touch my work and it would have less impact.
And you are confused. The immediate goal/purpose of academic work is to create knowledge. I said nothing about the underlying purpose of this goal. The ultimate purpose of this goal is to provide the foundation for a better society, which includes the knowledge being used to build better technologies, products, and processes by for profit companies.
The antipathy towards profit is one of the many great contradictions of the Left. They will willingly make the public worse off in order to prevent a private profit.
> And you are confused.
You literally said "The whole purpose of being an academic is to publish and disseminate knowledge." You are now saying this means something different than it's apparent reading?
> which includes the knowledge being used to build better technologies, products, and processes by for profit companies.
Why then do you think these entities are entitled to the output of academics without paying for it, or as a compromise, contributing back to the pool of global knowledge in kind?
What I'm getting at is the way you and the GP have framed this position is that you might then consider it wrong for an academic to charge for access to their work or for themselves to start a business and profit from it. That seems off to me.
> The antipathy towards profit is one of the many great contradictions of the Left.
This is an oversimplification of an easily understood nuance. For profit companies are obviously fine but we might also want a measure to determine if any one entity is contributing more to society than it is extracting from it.
In lieu of this perhaps forcing companies to contribute back to the same body of public knowledge that they obviously benefit from is not some great contradictory injustice.
> They will willingly make the public worse off in order to prevent a private profit.
Since business have no natural right to exist we, as people, have every right to exclude some of them from operating or even existing at all. Is my aversion to slavery actually a disguised antipathy towards profit?
And yet, without something like the GPL, you can be sure that the majority of improvements would eventually be sealed off behind doors and high walls. The whole reason for-profits wouldn't want to touch your work is because they would make _less_ profit, not _no_ profit. The real impact, as we see now (such as in BigPharma[0]), is things being more expensive than they need be with cheap alternatives being actively suppressed.
[0] https://www.aarp.org/politics-society/advocacy/info-2019/gen...
>They will willingly make the public worse off in order to prevent a private profit.
Because historically speaking, private profit does not fundamentally "provide the foundation for a better society". If we had more benevolent actors in my lifetime, I'd feel differently. But it's always a monkey's paw. I see no contradiction to be afraid of being exploited when everytime a cool idea scales massively it ends up exploiting me.
I’m kind of ok with this? I’ve written and had book chapters and research articles published. I never thought I was in any sort of position to restrict access. Publishing is about getting it out there. Attribution might be an issue, but that is a separate conversation and perhaps dealt with, if possible, by having LLM’s cite sources more accurately.
I have not kept up with the latest on LLM’s and licensing, but I’m curious: are scientific papers accessible to LLMs? Honestly, a bigger societal loss in my view is publishers like Elsevier restricting LLM access to research articles, rather than being too permissive. I could not care less if Elsevier makes a little bit of money in the process.
>Attribution might be an issue, but that is a separate conversation and perhaps dealt with, if possible, by having LLM’s cite sources more accurately.
hand-waving it by "oh just have corporations promise they will cite their sources, honor system" is the easiest way to not have corporations cite their sources. Plus if these corporations are so gun-ho about how "we don't store your data in an LLM", that wouldn't be a trivial matter anyway.
>a bigger societal loss in my view is publishers like Elsevier restricting LLM access to research articles, rather than being too permissive.
it's only a bigger loss if you think the knowledge will be proportionally handed down to society. Which it rarely is.
Aren't they also one of the academic publishers that's been criticized for charging for access to articles, and the authors don't get anything from the publication/distribution?
Which is why I have never paid for an academic article from a publisher.
I have been e-mailing authors directly for access to the article and have been successful most of the time. Maybe once or twice, the e-mails go unanswered. In one case, the primary author had died (a subsequent search linked a news article showing a skiing accident). In another case, the author had simply moved to a different university thus the e-mail in the paper was outdated.
This is pretty much a standard model in academic publishing
Except for open access models where someone -- usually the authors, but sometimes universities or other institutions -- pay so that the article can be accessed by the public for free. (Authors usually don't pay out of pocket but using grants, and OA journals usually offer some sort of "financial aid" for those unable to pay.)
What a waste of money to pay these parasites to let the public see your research.
It’s just as much a waste of grant money too. Whether that comes from or from a nonprofit, padding the pockets of some publishing company gatekeeper is in no way what anyone should expect their money to go toward.
True, and unfortunate, other than arxiv, and openreview
Except for the few open access journals out there, this is pretty much how all academic journals work.
It’s a little hard to tell from the article, but I think this is about books not research articles, and that the authors have some amount of revenue sharing.
Does publish mean something else in this context? I thought publishing means that anyone can have access to your research.
Think "publish" in the 1940's magazine sense - which is pretty much the distribution model for all "traditional" academic journals. If Famous_Author publishes an article in The Saturday Evening Post - maybe you're a subscriber to The Saturday Evening Post. Or you head to a brick-and-mortar library which subscribes, to read the article there. Or you pay Cover_Price, in cash, to a news stand or drug store or whatever, to get your dead-tree copy of the issue in question.
Anybody willing to pay for it, yes. Typically tens of dollars per article or thousands/year for a subscription. It's not like scientists get any of that money though, it's just one more case of publishers squeezing scientists.
Is there a legal way to bypass those collectors. It should be, right. After all most research is done on public money.
Email the author of the paper and ask for a copy. My success rate with that method is 100%.
“Greetings,
I am interested in reading your paper, “[title of paper],” but I have only been able to find paywalled versions. Would you please send me a copy?
Thank you,
trogdor”
Back in the day, this was the way to get really nice print copies by mail, and a handwritten letter by the author.
Finally! We've solved the problem of having to pay $30/article to get access. We'll just query the LLM!
can’t use that to develop AI, it’s against the terms of use
The whole academic publishing system is rotten to its core. Rent seekers living off the labour of underpaid academics and selling the product back to them, all while tax-payer money props up the racket.
Has anyone made a marketplace to sell data to AI and at the same time explicitly exclude data from being accessed by AI?
something like this already exists. they are called "data brokers"
Pretty disappointed by the responses here, but I suppose I can't be surprised in a community sympathetic to AI and abhorrent to copyright as a concept.
Those are both topics that can be a post in and of itself, so I'll just keep it simple and emphasize once again that we should implement the 3C's when asking of anything from another person's IP. I doubt many of the older papers/articles had contracts that allowed for such usage. Reinforced by the article:
>The agreement with Microsoft was included in a trading update by the publisher’s parent company in May this year. However, academics published by the group claim they have not been told about the AI deal, were not given the opportunity to opt out and are receiving no extra payment for the use of their research by the tech company.
regardless of your position, this publishing group at worst lied and at best is being irresponsible, this isn't even an issue of AI or copyright. We can debate "well this is how it should be", but let's leave ShouldLand for a bit and actually look at the current situation. Trust being broken in real time.
This is a rotten thing to do by Taylor & Francis. Humans are treated as expendable pawns to serve the capital and the machine.
We need new publishing models with strict copyright protections that protect against theft. Academics should run their own publishing houses as a cooperative.
The entire purpose of academic publishing is to widely distribute knowledge. The co-op would break doen the absolutely awful publishing fees to be sure, but… I’m not sure what “theft” you feel took place here.
And the distributors need to be rich mega-corporations who collect a fee for selling you back your own works?
They're not selling you back your works, they're selling you back a distillation of them. Hopefully one that's useful for a particular context.
great, my favorite aspect of society. Paying to get someone or something else's interpretation of raw knowledge that is barred from the public , privy to all their biases and in a black box resistant to peer review.
Yeah, this is definitely going to change the world. I'm not sure if it'll be for the better.
Who pays for the peer review? Who prevents the publication of nonsense, pseudo-science, cherry-picked findings, etc.
Granted you may argue that is done poorly today but it certainly doesn't happen for no costs.
You do realize peer reviewers are also unpaid.
No I didn't realize that. Why does anyone do them, and what real care do they take when they are doing it for free?
So you get a few benefits from being a reviewer but it's pretty limited.
One benefit is that you generally get free access to the journal in question for a month or two while you are reviewing. So if you are regularly reviewing papers, you get free access to that journal. If you are part of a big org you probably already have access but it is a nice little benefit for people outside big academic orgs.
You can also include it on your CV if you want and there are a variety of small recognitions you can get for doing it.
And some publishers will also give you a fairly substantial discount (25+%) on texts purchased through that publisher as well as similarly substantial discounts on publishing and editing services provided by the publisher. Sage does this for example and their services include things like translation services, editing, infographic & artwork design, and animating video summaries for papers. They seem like they are quite nice services so getting discounts on them would probably be a good selling point for some people.
Another reason is just because you'd be reading the journal anyways and you don't mind doing a deep dive every once in a while on a random paper in your field. This reason is mostly a "well why not I'm already spending the time" kind of reason but it's a decent justification.
And the main reason is just because it's kind of expected of a lot of PhDs and being good at reviewing is a useful skill to have, especially if you are going to eventually be someone else's advisor or just in general if you want to be able to effectively critique your own works and the works of those around you. Being a reviewer for at least some times means you know the system and you can help keep your own papers and the papers of your colleagues from getting rejected for dumb reasons or oversights.
> Why does anyone do them,
It's part of the culture. I write a paper. I want to get it published. Journals need someone to review it. If I refuse to peer review others' papers, then in principle people may refuse to review mine and we both lose.
Sure, it'd be nice if publishers passed some of the profits to both authors and reviewers, but that would create other perverse incentives.
> and what real care do they take when they are doing it for free?
Not much. They sometimes sit on it for months before reviewing it.
Pardon my ignorance, I’m very much not an academic, but what the heck do the publishers bring to the table here? Suppose you just published your paper on a free site, and solicited your peer reviewers to annotate a Google Doc or whatever? I can see why journals and their publishers mattered 50 years ago when they were needed to physically publish the information by printing it on paper and distributing it. But I don’t get it now. Why does anyone gift them their papers? To me this sounds like a store where the customers bring in all the merchandise and give it to the store, who then sells it back to other customers. In other words, crazy.
> but what the heck do the publishers bring to the table here?
Name recognition. Top journals are harder to publish into - you (supposedly) need a higher impact piece of work to get published in it.
Same idea with universities. The top ranked universities don't necessarily give you a better education. But that certificate sure helps.
> But I don’t get it now. Why does anyone gift them their papers?
Same answer as above. You're a researcher who is trying to get tenure. You published in Nature. Good chance you'll get tenure. You published on your own site and have a Google Doc of reviewer feedback? Anyone can create that. You won't get tenure.
At least in my field, being on the program committee for top conferences is prestigious and can help with promotion, increases your visibility and profile in the community, is a good networking opportunity with other experts in the field, and for early career researchers provides a valuable perspective on how other experienced researchers evaluate research. The latter is less important in cases where conferences publish all reviews.
Who would pay them, and would you trust their review if they were paid?
> Who prevents the publication of nonsense, pseudo-science, cherry-picked findings, etc.
no one, all of that gets published constantly
> We need new publishing models with strict copyright protections that protect against theft.
No, we need new publishing models that release the publicly-funded information into the public domain. If tax dollars paid for the research then there should be no copyright on the results of that research at all, owned by the publisher or authors.
This wouldn't represent a change for the authors anyway, who already are accustomed to having to sign away copyright (hence this story). The problem in this instance isn't that the papers are being fed into machine learning models, the problem is that the publisher is extracting a check for the privilege.
If no one is going to learn from academic papers/data what’s the point in doing it? People care less and less about the impact more about how much money it’ll make them. If that’s all you want from it then be honest but don’t complain when someone makes more money from something than you.
The point of academic papers is that HUMANS learn from it. The point is not that machines generate a mashup that masquerades as a probability distribution and companies sell that mashup together with ads (the latest predictable development).
I see no indication that the affected academics are concerned with the money that their papers make them. Most want to give them away for free to HUMANS.
Excellent. More humans will have access to academic paper information in a form that's useful to them now :)
In this age of mass misinformation and astroturfing, I guess it would be more "useful" to them now. No one seems to care about the truth these days.
surprised_pikachu.jpg
We need regulations that enforce positive consent. Otherwise, this will keep happening.
We need to eliminate the rent seeking behavior of academic publishers. They do not fund the research, do not fund the peer review of the research, but pocket huge profit margins by charging others[1] for access to it.
Edit: [1] By others, I mean our university libraries and any poor soul who does not have membership in one. Today, universities pay to access their own research.
We need to eliminate the publishers period. Publishing on the Internet is free.
Publishers do a lot more than print things out. I'm not saying that what they do is good, but "let's just put it on the internet" doesn't replace publishers any more than rideshares replace long-haul trucking.
What, exactly, is this "a lot more"? I am an academic who publishes regularly, and I'm having difficulty figuring this out.
They clearly employ lots of bureaucrats, middle managers, and lawyers. They are constantly trying to make sketchy side deals with third parties, as T&F did. They employ copyeditors, who in theory (and sometimes in practice) improve papers but who very often are totally incompetent. They maintain large databases of something-or-other, presumably so they can sell out my data. They maintain complicated systems to submit and review papers, when it seems that email would suffice, presumably so they can track all this metadata they want to sell. Oh! And they lobby the government for their own selfish ends:
https://www.theguardian.com/science/political-science/2018/j...
Seriously. Imagine all of them up and vanished tomorrow. Would anybody miss them?
Again, I didn't say that what they do is the right way to do it, but what about discoverability? Would researchers just submit their document to every library in the world? If no entity is vetting these papers in a basic way, what's to prevent the entire channel from being taken over with spam? Is it done entirely by URL, and what's to stop the cited content from changing? What if they lose their domain? Is the similar domain that pops up with similar content the same author or someone that just ripped the site and is going to start serving up ads as soon as everyone switches their URIs? Or maybe we'd have libraries go out and actively retrieve this stuff rather than passively accepting it-- Would they monitor... mailing lists? What if someone posts a timely and incredibly controversial paper that lists a respected authority that just died as its author? What's to stop an author that published a bullshit paper from passively claiming it was a forgery? Would you, currently, give equal weight to citations published in the New England Journal of Medicine and a google doc with a bunch of citations to other google docs and personal websites? If we have centralized locations for this work, how would they not be publishers? Beyond that, current URL citations are a fucking mess. Link rot is a huge problem and most of the people that make citations in papers have no idea how to mitigate it.
Discoverability, relatively stable hosting, and the most base-level vetting and credibility are problems that they address, even if they do so poorly and greedily. If they went away tomorrow, so would the databases of work they manage, the APIs libraries use to query them, their mechanisms to confirm that whoever said they published something actually published it... Publishers don't have to be the answer-- I think a lot of these problems could be resolved through field organizations, librarianship and university organizations, but what's the likelihood that of universities and field organizations will try to monetize it and essentially become publishers in all of the worst ways?
> what about discoverability?
That's what Google and other search engines are for. arxiv.org even has its own search function. Nobody depends on journals for this.
> Would researchers just submit their document to every library in the world?
No, just post it on a known website like arxiv.org. Then anyone with Internet access can get to it, and they don't even have to go to a library.
> If no entity is vetting these papers in a basic way
You do have to have academic credentials to post papers on arxiv.org.
Not only that, but plenty of papers published in journals turn out to have been bogus--not just wrong (many papers published in good faith later turn out to be wrong, that's part of science) but based on faked data, and nobody in the journal review chain spotted it. So the supposed vetting that journals provide doesn't really add any value anyway.
> What if they lose their domain?
arxiv.org is supported by many academic institutions and foundations. They're not going anywhere.
> If we have centralized locations for this work, how would they not be publishers?
By being arxiv.org or something like it. In other words, there is already an existing example of a non-publisher that addresses all the problems you say journal publishers address--and does it much better.
The Internet Archive et al alleviates the majority of those concerns. And it isn't publishers in and of themselves that make a work credible, but the peer reviewers. I can easily see some extra function on IA that allows verified users, some of whom will be peers, to formally review and endorse works. Want to find high quality works? Search and filter by number of endorsing reviewers with related works on their profile.
> the most base-level vetting and credibility are problems that they address
No they don't. They get that for free too, with "volunteer editors".
Stable hosting and gatekeeping, that's about it.
Stable hosting--which we, the taxpayers, who already paid for the research that was published, then have to pay again to get access to, unless we're affiliated with an academic institution, in which case the academic institution is paying for access to papers that the taxpayers already paid for.
Even for the large amount of research that isn't publicly funded and is done by private research organizations and universities, there's no reason it shouldn't be available to the general public, for free. I wish someone would actually do it while meeting the use cases of the other stakeholders.
They’re branding experts. Getting published by Nature is far more valuable to someone compared to (all the other options). Humans prize exclusivity. We’ll hand our time and resources to people who can give us a tiny bit of it. The harder we have to work, the better.
A fancy watch. A low production supercar you can’t buy unless you have 5 of the same marque already. All the same broken stupid pattern.
> Seriously. Imagine all of them up and vanished tomorrow. Would anybody miss them?
Probably not much. Arxiv works pretty well. Usenix is already open access. NSF-funded research already has to comply with their 12 month open access policy (zero months would be better though.)
I don't see many good reasons for journals to stick with non open access publishers.
If what they do isn't good, why would we want them to keep doing it?
As an academician I wholeheartedly support this. Some journals are now charging thousands of dollars to make our work open-access. This is an industry run by sociopaths.
Publicly funded science shouldn't be "published" by some corporation, it should go directly into the public domain.
Having a totally artificial middleman makes no sense when basically all publishing happens digitally. And the public deserves full access to the science they have funded.
That would be the (oddly named) green open access model.
https://en.wikipedia.org/wiki/Open_access#Colour_naming_syst...
The reality is that AI publishers (NeurIPS, ACL, related) all do proper academic publishing norms, and they are now forcing the rest of the world to follow them or have their content laundered in the form of LLMs trained on it. Good.
Why is a researcher called a creator in this article?
Are they not a fact discoverer or truth revealer?
It's unclear to me researchers should “own” truths prior research and public patronage enabled them to unearth.
// note: research != invention, i.e., Space X experimenting until systems and machinery can land a rocket on a barge is not “research”, but testing and documenting characteristics of fuels in a vacuum as the environment swings from -100C to 120C is
>Space X experimenting until they can land a rocket on a barge is not “research
Why? What else could it be if not research?
> what else could it be if not research?
it could be experimentation and learning about your own invented system and mechanic, not uncovering of existing truths
the word "search" makes this more clear, the answers research finds already exist, we just didn't find them yet. that is not the case for inventions that don't work, and need to be experimented with until they work.
there may, certainly, be some incidental research discovery while attempting to make an invention work (e.g. if Space X had hit on something new about physics itself that had been causing their prior experiments to fail) -- but nobody "owns" such truth discovered, while of course they own their invention that works as a result of applying known truths to a creation.
The usual pairing is research and _development_ isn't it.
I’m not sure why you’re being downvoted, it is a valid and important point. Researchers discover empirical structure. They don’t create it. It isn’t art.
The article seems to want to bring rights holders together in a way that isn’t valid.
A scientific paper is normative. Its construction is closer to a pull request than artwork.
Seems like an unnecessary nitpick, which is discouraged in HN. you can argue the opposite way and say artists "discover empirical structure" that resonates with a lot of people. But colloquially we call it an "art style", often named after the "discoverer".
Interesting inverted/remapped take.
The form is what is under copyright, and the form of the paper is normative, while I would argue the form of the art, if normative, makes it illustration and not art.
If I change the hypothesis and data of a paper I change the entire meaning and consequence of the paper, although the style remains constant. If I change the palette and subject of a painting, the style remains the same and it would not be considered anything more than trivially novel.
If style is art, then there is no art in scientific papers.