Mostrando entradas con la etiqueta Inteligencia artificial generativa. Mostrar todas las entradas
Mostrando entradas con la etiqueta Inteligencia artificial generativa. Mostrar todas las entradas

lunes, 18 de mayo de 2026

ELSEVIER vs META-Zuckerberg: ladrón que roba a ladrón... o la defensa del negocio de los artículos científicos

Publicado en THE Times Higher Education
https://www.timeshighereducation.com/news/open-season-scholars-rights-if-elsevier-loses-meta-fight



«Se abrirá la veda» sobre los derechos de los investigadores si Elsevier pierde la batalla contra Meta


Según los expertos, una derrota del gigante editorial en su demanda judicial sobre el rastreo mediante IA dificultará mucho más que los investigadores puedan controlar cómo se utilizan sus trabajos


Publicado el 15 de mayo de 2026


Mientras que la editorial académica más grande del mundo se ha enfrentado a boicots y a dimisiones masivas de editores de revistas debido a las preocupaciones sobre sus precios y márgenes de beneficio, el anuncio de Elsevier de que se une a una demanda colectiva junto con otras cuatro editoriales importantes para demandar a Meta y a su fundador, Mark Zuckerberg, por «infracción deliberada de millones de obras textuales… para desarrollar los modelos de lenguaje a gran escala (LLM) Llama de Meta» podría atraer un apoyo significativo de la comunidad académica, según afirmó Alina Trapova, profesora de Derecho de la Propiedad Intelectual en la Facultad de Derecho de la UCL. 


«Puede que los académicos no hayan considerado a Elsevier como la editorial más favorable a lo largo de los años, pero ahora parece que está luchando por los derechos de los autores», declaró Trapova a Times Higher Education con motivo de la primera demanda relacionada con la inteligencia artificial (IA) presentada por varias editoriales líderes.


Alegaban que Meta había recopilado millones de artículos y libros pirateados ilegalmente para entrenar su modelo de IA generativa, el cual, a su vez, había devuelto a Internet unos 40 terabytes de datos, equivalentes a unos 5 millones de libros. Muchos de los resúmenes de Llama sobre resultados de investigación contienen alucinaciones e inexactitudes que podrían dañar la credibilidad de los académicos, afirma Elsevier. 


«Para los académicos, obtener reconocimiento a través de la atribución es realmente importante. Necesitamos que esas atribuciones nos lleguen a nosotros, en lugar de que nuestro trabajo acabe convertido en resúmenes anónimos generados por IA o en una simple alucinación», afirmó Trapova al explicar por qué los académicos se pondrán del lado de Elsevier.


Meta alega que el uso que hace de los artículos para entrenar su LLM constituye un «uso legítimo», pero, de aceptarse, esto ignoraría los derechos de los autores sobre sus propios materiales, que Elsevier ha tratado de gestionar mediante acuerdos individuales, continuó Trapova.   


«Se ha pedido a los colaboradores que firmen estos documentos indicando si permiten que su trabajo se utilice en diferentes contextos; Meta ha interpretado este acuerdo en el sentido de que “podemos hacer lo que queramos con este trabajo”, lo cual es una maniobra realmente desagradable», afirmó.


En este sentido, «está en juego la integridad académica», aunque «este no es el aspecto central de la demanda», continuó Trapova.


«Para Elsevier, esta demanda tiene como objetivo principal proteger su modelo de negocio, y los derechos de autor son el mecanismo legal para hacerlo. Pero también velará por los intereses de los autores y creadores», afirmó.


Aunque a muchos académicos les resulte difícil simpatizar con Elsevier —cuya empresa matriz, RELX, tiene una capitalización bursátil de 43 000 millones de libras—, la defensa jurídica de sus intereses económicos podría contribuir a reforzar el control que los autores tienen sobre sus obras y el uso que se hace de ellas, señaló Trapova. «El caso no se centra principalmente en la integridad de la investigación: el marco jurídico de EE. UU. tiene un aspecto moral, pero las protecciones en este ámbito son mucho más débiles. Por eso, lo que realmente se debate en los tribunales gira en torno a la pérdida económica», afirmó.


Emily Hudson, profesora de Derecho en la Universidad de Oxford, cuya investigación se centra en los derechos de propiedad intelectual, coincidió en que los académicos se pondrán en su mayoría del lado de Elsevier.



«Hay mucho apoyo... a las acciones legales contra las empresas tecnológicas», explicó Hudson, y añadió: «Esto no quiere decir que los académicos estén en contra del uso y el desarrollo de la IA en sí. Más bien, creo que la preocupación se refiere a cómo se está utilizando la IA».


«Si un investigador quiere utilizar mis artículos para, por ejemplo, desarrollar una IA que ayude a diagnosticar enfermedades, entonces me parecería bien que lo hiciera de forma gratuita y, de hecho, sin pedirme permiso. Por otro lado, me sentiría mucho menos cómoda si una empresa tecnológica utilizara el mismo conjunto de artículos para desarrollar un producto de IA generativa que ayude a los usuarios a redactar correos electrónicos o escribir ensayos», explicó.


«Dicho esto, debemos tener claro que los intereses de las editoriales y los académicos no coinciden del todo», afirmó Hudson, señalando que un acuerdo de compromiso podría ser aceptable para Elsevier, pero no para los autores.


«En este caso, dado que son las editoriales las que impulsan el asunto, es posible que el resultado final que ellas prefieren o consideran aceptable (como acuerdos de licencia con empresas tecnológicas) no sea el preferido por los autores académicos. Así pues, aunque tanto las editoriales como los autores puedan tener interés en demostrar que las empresas han infringido los derechos de autor en el desarrollo actual de la IA, puede que no haya acuerdo sobre cómo se debería gestionar esto en el futuro». 


Sin embargo, Caroline Ball, responsable de participación comunitaria en Open Book Collective, una organización que defiende la publicación de acceso abierto, afirmó que el enfrentamiento entre Elsevier y Meta era «complicado» para los académicos.


«Es más complicado que el caso de piratería de libros de Anthropic, ya que en aquel caso los derechos de los autores eran el núcleo del asunto, y cualquier compensación económica iría destinada principalmente a los autores», afirmó Ball en referencia al acuerdo de 1100 millones de libras esterlinas alcanzado en septiembre de 2025. 


«Si Elsevier ganara y obtuviera una indemnización o compensación, es poco probable que nada de eso llegara a las instituciones o a los académicos, ya que, en la mayoría de los casos, estos habrán cedido sus derechos a Elsevier en el momento de la publicación. Así que solo supondrá más dinero para Elsevier a costa del trabajo académico de estos», afirmó.


«Si Meta ganara y su uso generalizado de material protegido por derechos de autor se considerara legal, o lo que es más probable en los tribunales estadounidenses, “uso legítimo”, entonces se abriría la veda sobre el material protegido por derechos de autor», continuó Ball, añadiendo: «¡Sospecho que para la mayoría de nosotros en la comunidad académica se trata de un caso de “una plaga para ambas partes”!».


*****************************************

Open season’ on scholars’ rights if Elsevier loses Meta fight


Loss for the publishing giant in its legal action over AI crawling will make it much harder for scholars to assert control on how their outputs are used, say experts


Published on May 15, 2026

Last updated May 15, 2026

Jack Grove

Twitter: @jgro_the



While the world’s largest academic publisher has faced boycotts and mass walkouts by journal editors over concerns over its prices and profit margins, Elsevier’s announcement that it is joining a class action with four other major publishers to sue Meta and its founder Mark Zuckerberg over “wilful infringement of millions of textual works…to develop Meta’s Llama large language models (LLM)” may attract significant support from the scholarly community, said Alina Trapova, lecturer in intellectual property law at UCL’s Faculty of Laws.  

“Academics might not see Elsevier as the friendliest publisher over the years but now it seems it is fighting for the rights of authors,” Trapova told Times Higher Education on the first artificial intelligence (AI) action brought by several leading publishing houses.

They allege Meta scraped millions of illegally pirated articles and books to train its generative AI model, which has, in turn, returned about 40 terabytes of data, equivalent to about 5 million books, to the internet. Many of the Lllama summaries of research outputs contain hallucinations and inaccuracies which could potentially damage the credibility of academics, claims Elsevier.  

“For academics, getting recognition through attribution is really important for us. We need those attributions coming to us rather than having our work spat out in anonymous AI summaries or a hallucination,” said Trapova on why academics will side with Elsevier.

Meta claims its use of articles for training its LLM represents “fair use” but, if accepted, this would ignore authors’ rights to their own materials which Elsevier had sought to manage via individual agreements, continued Trapova.   

“Contributors have been asked to sign these documents stating whether they permit their work to be used in different contexts – Meta has interpreted this agreement as that ‘we can do anything with this work’ which is a really nasty move,” she said.

In this respect, “academic integrity is at stake”, although “this is not the central piece of the claim”, Trapova continued.

“For Elsevier this lawsuit is about mainly about protecting their business model and copyright is the legal mechanism to do this. But it will also be looking after the interests of authors and creators,” she said.  

While many academics may find it hard to sympathise with Elsevier, whose owner RELX has a market capitalisation of £43 billion, its legal defence of its economic interests may help to uphold how authors can control their works and how they are used, said Trapova. “The case isn’t primarily about research integrity – the legal framework in the US does have a moral side but protections on this side are much weaker. That is why what is actually in court is framed around economic loss,” she said.

Emily Hudson, professor of law at the University of Oxford, whose research centres on intellectual property rights, agreed that academics will mostly side with Elsevier. 

“There is a lot of support...for legal actions against tech companies,” explained Hudson, adding: “This is not to suggest that academics are against the use and development of AI, per se. Rather, I think the concern relates to how AI is being used.”

“If a researcher wants to train on my papers to, say, develop an AI that helps diagnose diseases, then I may be happy for them to do so for free, and indeed without asking me for permission. On the other hand, I may be far less comfortable with a tech firm training on the same set of papers to develop a generative AI product that helps users draft emails or write essays,” she explained. 

“That said, we need to be clear that the interests of publishers and academics do not completely align,” said Hudson, noting that a compromise deal might be acceptable to Elsevier but not authors.

“In this case, because we have publishers driving the case, it may be that their preferred or acceptable endpoint (such as licensing deals with tech firms) may not be that preferred by academic authors. So while both publishers and authors may have an interest in showing that firms have infringed copyright in existing AI development, there may not be agreement on how this should be handled in the future.”

However, Caroline Ball, community engagement lead at the Open Book Collective, which champions open access publishing, said the Elsevier-Meta face-off was a “complicated one” for academics.

“It’s more complicated than the Anthropic book piracy case, since in that case authors’ rights were at the heart of the issue, and any financial recompense would go to the authors, mostly,” said Ball on the £1.1 billion settlement agreed in September 2025.

“Should Elsevier win and get a payout or compensation, it’s unlikely any of that would make its way back to institutions or academics, since in most cases they will have signed their rights over to Elsevier at the point of publication. So it will be just more money for Elsevier on the basis of their academic labour,” she said.

“Should Meta win, and their wholesale use of copyright material be deemed legal, or more likely in US courts, ‘fair use’, then it’s open season on copyrighted material,” Ball continued, adding: “I suspect for most of us in the academic community it’s a case of ‘a plague on both your houses’!”

jack.grove@timeshighereducation.com


miércoles, 28 de enero de 2026

In Memoriam: The Academic Journal / ¿Ha muerto la revista académica? Un obituario crítico del sistema de publicación científica

¿Ha muerto la revista académica? Un obituario crítico del sistema de publicación científica


Un texto ensayístico firmado por Russell Beale, académico de la University of Birmingham, propone una provocadora idea: la revista académica tradicional ha llegado al final de su vida. En forma de obituario, el autor recorre más de tres siglos de historia editorial —desde el Journal des Sçavans (1665) hasta la era de la inteligencia artificial generativa— para argumentar que el sistema se desvió de sus valores fundacionales y terminó colapsando bajo sus propias lógicas de mercado, métricas y automatización.

https://arxiv.org/html/2512.23915v1

In Memoriam: The Academic Journal
Russell BealeR.Beale@bham.ac.uk
School of Computer Science, University of Birmingham, Edgbaston,
Birmingham, B15 2TT, UK
(c) 2025 IEEE. All rights reserved, including rights for text and data mining, and training of artificial intelligence and similar technologies.
Abstract

In this piece we reflect on the life and influence of AJ, the academic journal, charting their history and contributions to science, discussing how their influence changed society and how, in death, they will be mourned for what they once stood for but for which, in the end, they had moved so far from that they will less missed than they might have been.

martes, 20 de enero de 2026

¿Un pez en artículos médicos? El extraño rastro de textos “contaminados” en revistas científicas

Publicado en no break throughs
https://nobreakthroughs.substack.com/p/is-this-fish-a-sign-of-fraud-fabrication




¿Un pez en artículos médicos? El extraño rastro de textos “contaminados” en revistas científicas 

En los últimos meses, investigadores y editores han comenzado a detectar algo desconcertante en artículos científicos revisados por pares: la aparición de nombres latinos de especies animales y vegetales —peces, plantas e insectos— insertados sin sentido en investigaciones sobre salud humana, ingeniería o ciencias sociales. El fenómeno no responde a una disciplina concreta ni a un solo editor, y por ahora no tiene una explicación clara.

* El hallazgo inicial: el fenómeno fue señalado por Leon Xiao, profesor en la City University of Hong Kong, al detectar nombres científicos como Parazacco spilurus en artículos sin relación alguna con biología.

* Textos fuera de lugar: especies animales y vegetales aparecen en estudios sobre fragilidad cognitiva, lesiones en bomberos, biomarcadores o pérdida auditiva, sin justificación científica.

* Editoriales afectadas: los artículos identificados pertenecen a revistas de grandes editoriales como Springer Nature, Elsevier y Frontiers.

* Patrones geográficos: hasta ahora, los artículos con estas anomalías tienen autores afiliados a instituciones chinas, aunque no se establece una causa directa.

* Relación con “tortured phrases”: el fenómeno recuerda a las llamadas tortured phrases, sustituciones absurdas de términos científicos detectadas previamente en miles de artículos y asociadas a traducción automática o parafraseo algorítmico.

* Ejemplo documentado: en un artículo de Scientific Reports sobre lesiones en bomberos, Parazacco spilurus aparece repetidamente en tablas estadísticas y nombres de variables.

* Respuesta editorial: el editor en jefe de Scientific Reports confirmó que se investigará tanto el artículo señalado como el problema de redacción a mayor escala.

* Más allá de la ciencia: los mismos nombres latinos aparecen en novelas web y listados comerciales, lo que sugiere el uso de herramientas automatizadas compartidas entre contextos no académicos y científicos.

* Sin conclusión definitiva: por ahora no se puede afirmar si se trata de fraude, errores de traducción, uso de IA o una combinación de factores, pero el paso de estos textos por la revisión por pares genera preocupación.


***************************

Fishy Text in Research Articles: Fraud or Fault?

A reel big mystery as a strange new species of error appears in major journals

Jan 11, 2026





What do the predaceous chub (Parazacco spilurus), a non-existent species of rattlebox moth (Utetheisa kong), and a paper mulberry plant (Broussonetia papyrifera) have in common?

They’ve all started appearing in scientific journal articles over the last few months, but peculiarly, you can find them in articles unrelated to the study of wildlife biology or conservation or botany. The fauna and flora are appearing in research about hearing loss in the elderly, nerve repair and ankle injuries in firefighters. Wha….?

What does the predaceous chub have to do with that?


Last week, I received an email from Leon Xiao, assistant professor at the City University of Hong Kong. He had begun to notice Latin names for a variety of species were appearing in all sorts of places. After a little sleuthing, he’d discovered the phrases were turning up in scientific papers by various publishers, including Springer Nature, Elsevier and Frontiers.

Xiao wanted to bring it to the wider attention of the scientific community and I wanted to dig in and try and figure out exactly what is going on.

Spoiler alert: I haven’t figured it out yet. But have found a ton of Weird Stuff.

Of these phrases, the one Xiao was seeing regularly was:

Parazacco spilurus subsp. spilurus.1

This is the scientific name for the predaceous chub, though the “subsp. spilurus” is an unnecessary addition2. This isn’t an impressive fish. It’s like the vanilla icecream of fish. If someone showed you a picture of it, you’d say “that’s a fish”.

VIDEO : https://youtu.be/Z8wIx0xqXEE

But Parazacco spilurus is showing up in several journal articles with the following headlines:

  • “Integrated transcriptomic analysis identifies lysosomal autophagy-related genes in sarcopenia”

  • “Analysis of influencing factors of cognitive frailty in older adults community patients based on restricted cubic spline”

  • “Flow cytometry-based validation of soluble biomarker detection”

There is no reason why the chub should find itself in these places and, indeed, if it did it would not end well for the fish, I imagine.

Among the papers, there are not always obvious commonalities in content, but all authors from papers featuring Parazacco so far appear to be from Chinese institutions.

And it’s not only research where Parazacco is appearing.

It’s also, weirdly, finding itself in listings on Alibaba for … adhesive?

So, here we have ourselves a little mystery, a new species of unknown error in science…

Mysterium verbi distorti

In keeping with our Latin theming so far, I’m calling this phenomenon Mysterium verbi distorti.

I’ve taken it from a (probably extremely poorly translated3) idea of “Tortured Phrases” — phenomena in the scientific literature which have become a sign of potential error and integrity breaches.

If you’re not up with your scientific integrity investigations, tortured phrases are a concept identified by Guillaume Cabanac, Cyril Labbe and Alexander Magazinov a few years back. You don’t have to look far to find them. Cabanac and co. found them in more than 3,000 papers to January 2022.

They’re bloody good fun too. For instance, you may be so lucky to have seen the term “bosom peril” appear in a scientific paper. This is a rephrasing of the term “breast cancer”. Similar substitutions appear for words like artificial intelligence (replaced with counterfeit consciousness [lol]) or smart home (replaced with clever home [lmao]).

When Xiao pointed these Latin names out, tortured phrases were the first thing I thought of: Perhaps these substitutions were being made via automated tools — either AI or word libraries or paraphrasing tools — as non-English speaking authors were translating their work into English. The errors then, somehow, slipped through editing and the peer review process, which is a little embarrassing.

For instance, take this article, published in Springer Nature’s Scientific Reports on October 29, 2025:

Epidemiological attribution of knee and ankle injuries in firefighters

The study’s abstract says it “aimed to evaluate the main factors of knee and ankle injuries among Chinese firefighters, focusing on the role of training arrangements, training conditions, and knowledge of training injuries.”

And yet, here we find the predaceous chub, that devil of a fish!!!!!

In Table 7, Parazacco spilurus appears across the final row of the entire table. Every single entry begins with the Latin name.


The section that follows this table also includes another Latin species name — broussonetia papyrifera — that is inserted directly in the middle of another sentence.

This study is about human beings, sure, but there are three instances in this paper where Homo sapiens appears. Two of these appearances are bizarre. The paper says:

“Secondly, the included risk factors such as “training load (A2)”, “homo sapiens protective equipment (B4)”, and “rehabilitation measures (C4)”

It seems in the corresponding table that discusses statistics on these variables, Table 4, that the B4 category is titled “personal protective equipment”. This seems to me like a substitution, similar to that seen in tortured phrases above.

I reported the issues in this paper to Scientific Reports on Friday last week. The team provided a statement on Tuesday, January 13.4

It comes via Chief Editor, Rafal Marszalek: “Thank you for flagging this. We'll investigate the paper thoroughly, and my colleagues in the research integrity team will dig into the phrasing issue more generally and raise it with the wider publishing community.”

These aren’t the only terms. So far, here’s the list of strange Latin names that have been inserted across the literature:

  • Parazacco spilurus subsp. spilurus

  • Broussonetia papyrifera

  • Utetheisa kong

  • Phoxinus phoxinus subsp. phoxinus

  • Castanopsis chinensis


Some of the offending species, left to right, Broussonetia papyrifera, Utetheisa species (not kong), Phoxinus phoxinus and Castanopsis chinensis

Most of these species names seem to appear subsequent to publication in August. Some of the papers I have viewed were accepted, at the latest, around September. There are at least 20 papers identified with the first phrase, and likely more that I can’t access with a freelancers institutional login :|, the others are a little less common, though Broussonetia and Utetheisa are appearing a little more.

Of the above five, there’s another strange place where you will find all these Latin phrases…


On the website, Webnovel, I stumbled across this storyif we can be so generous to call it that:

The earth is frozen: I have built a doomsday safe zone

I didn’t read the whole thing because my brain is already polluted enough by the internet but my guess would be that it’s AI-generated.

Several chapters (all?) feature Latin species names totally unrelated to the story. For instance, Chapter 12 includes both Parazacco spilurus subsp. spilurus and Utetheisa kong, in amongst a story about, I assume, the guy from I Am Legend trying to store water in his warehouse at the end of the world.

You can see, it also features weird additions.

Phoxinus phoxinus, dear reader, is a Western European fish colloquially known as the “minnow”. It has, to my knowledge, not escaped the bounds of the marine world and started trading stocks and crypto.

SO what the HELL is going on here? We have scientific literature, strange doomsday webnovels and Alibaba express listings.


Well:

  1. I am not sure yet but I have decided to point this out so the scientific publishers and researchers with peer review assessments may look for these terms, or other strange terms, in the literature right now.


    [Small update 12-01: some speculation by followers on Bluesky using LLM that this is a translation tool issue. This may well be the case and there seems to be evidence for it, but that doesn’t really give us a reason as to why these phrases made it into so much of the scientific literature…]5

  2. I have had contact with several authors of the papers here. The responses have not shed light on how these insertions came to be. I believe one response was generated by AI and the documents provided by the researchers as evidence of an AutoText mess up seemed to have been created in response to the discovery. I am continuing to dig into this.

  3. At the very least, the spurious and nonsensical word additions should be removed from these papers and the publishers should investigate where the data associated with these papers comes from.

  4. In terms of timing, all the spurious references seem to appear in the literature after August 2025, though some articles were submitted around May. Does this correspond with some firmware update of a particular LLM, AI or translation software?

  5. The fact all of these terms appear in a webnovel with well over 70 chapters could be suggestive of the fact they were generated, perhaps in Chinese, then translated with a tool, which did not accurately convert across. Whatever tool is being used, it’s apparent that the tool is in use in the scientific world, and may be able to be accessed by the public. That’s a useful thread to pull on.

For now, our mysterium verbi distorti remains just that. It’s too early to say just what these phrases might indicate, but alarm bells are ringing!!!! If it’s merely a translational error, then that still makes me feel a little uneasy — so much of this stuff is getting past peer review, into high quality journals

I hope to have a little more detail next week…



ELSEVIER vs META-Zuckerberg: ladrón que roba a ladrón... o la defensa del negocio de los artículos científicos

Publicado en THE Times Higher Education https://www.timeshighereducation.com/news/open-season-scholars-rights-if-elsevier-loses-meta-fight «...