📬 Interaccions #061 Sobre Internet Archive i la conservació de la memòria digital: molt més que una web, és la biblioteca universal del nostre temps. 👉 www.interaccions.cat/p/interaccio...

Amb el fantàstic equip de @radiowebmacba.bsky.social vam estar xerrant hores i hores, 2 dies separats en el temps... Em van donar de menjar i em van fer deixar pensar i perdre'm en bones mans.

But Glenn! It's $1.5 BILLION dollars in "licensing." You ARE aware that Anthropic's current valuation is $183 BILLION dollars, right? This isn't even a rounding error. It's legalizing IP theft. It's cheaper than licensing the works. It's a bargain for Anthropic.

Bé, no tothom cobrarà perquè s'està veient que alguns editors van "oblidar" registrar el copyright (en resposta sota aquesta cita). 3000€ o $ és misèria per un ús universal d'una obra en un xatbot, a ull de taxador. Equival a uns 1500 llibres venuts, posem. I de quants usuaris de Claude parlem?

i love this. Cursor is anthropic's biggest compute customer, and sends 100% of their revenue to Anthropic so that Anthropic can buy ads to sell Claude Code, a competitor to Cursor, which they are going to make better by buying data off of Cursor created by Cursor using Anthropic's models

Del blog estupendíssim de la Biblioteca de Catalunya, una entrada divulgativa fabulosa, de la Carolina Taboada, de Digitalització, sobre com es crea una col·lecció a la Memòria Digital de Catalunya, tot aprofitant l'ocasió de la incorporació del Fons Enric Granados. www.bnc.cat/El-Blog-de-l...

How does #WaybackMachine prevent intervention or corruption of archived information ?… seriously, the Trump Admin is wiping history out of the Smithsonian

The #WaybackMachine is more than nostalgia—it proves the web has a history. Sites vanish & links break, but memories & knowledge endure. That sense of safety & continuity is why preservation matters. #Wayback1T

"But a system designed to sanitize chatbot conversations does not serve the purpose of academic search. It actively undermines it." More info on this from Aaron Tay aarontay.substack.com/p/the-ai-pow...

[Read] Cognitive ease at a cost: LLMs reduce mental effort but compromise depth in student scientific inquiry www.sciencedirect.com/science/arti... - Fun read. Looking for more empirical studies comparing people using Search engine vs LLM.

The Brick House Cooperative also publishes @flaminghydra.com Our work supports digital ownership rights, cultural preservation, libraries and archives, and freedom of expression Try Flaming Hydra out here: flaminghydra.com/free

We’d love to know if people are using rescued data. We have download numbers but it would be helpful to know how people are using them! If you have use cases get in touch!

Have you submitted an idea for European Public Domain day? Deadline is Sept 31.

#STOPIAGENERATIVA Las Autoras de Cómic seguimos luchando por nuestros derechos de autoría y propiedad intelectual. Desde el Colectivo de Autoras de Cómic seguimos trabajando incansables en varias acciones, como la itinerancia de la exposición “Autoras de cómic en contra de la IA generativa”⏩

Them: “Cities are all rotting hellscapes!” Cities:

Esto resumen bien la mentalidad de tanta gente. Lo que les molesta de pagar impuestos no es pagar; es que sean impuestos. Prefieren pagar MÁS, pero que no sean impuestos.

Great that Ireland already has this for print, ebooks and audio books

The beautiful rare book display at the Beinecke Rare Book and Manuscript Library at Yale. We stopped by this week and had to share with you all.

Where do you think the U.S. version will be? Sign the petition against book bans at www.bookbanpetition.us

It's become quite a widely used term, but I don't like 'hallucinate' used for LLMs. They don't perceive the information. I would instead use 'confabulate', the plausible filling in of knowledge gaps.

My video about how LLMs are not search engines has led to many, MANY comments telling me that I should be using Perplexity. Some insisting that Perplexity does not hallucinate. Out of a list of 26 papers it just provided me (in "Research" mode) 4 were real. FOUR. 85% hallucination rate.

In 25 years, every business school in the country will be doing case studies about how a long defunct company known as “Google” once had an unbeatable lock on online information retrieval and then started doing shit like this.

Hypertext was a product of the 60s... because its spiritual and technical precursor, the Mundaneum institute's Universal Decimal Classification, was functionally destroyed by the Nazis razing Belgium in 1940.

How are we, in the year 2025, still dealing with library workers who want us to be meek, have no opinions, and act, essentially, as middleman robots that dispense whatever a patron asks for, whenever, with no guidance, no questions, no pushback?

To the person angry we weren't "neutral" in our AI Ethics panel: We absolutely should be saying no to technology that poisons the environment. We absolutely should be making moral choices about who we give money to. There is nothing neutral about librarianship. Never has been, never will be.

The fact that the thing we're calling artificial intelligence *can't do math* and yet we're jamming it into programs that successfully *have done math* for decades, then warning people against using the AI to do math, seems like an excellent summary of where we are.

Quien lo usa para chorradas ni se lo plantea ni le importa.

This is really cool! The Embassy of the Free Mind in Amsterdam is doing some really interesting stuff. Check them out.

“Vibes are not enough,” he said. “You’re not taking a good enough look at your real data. Is it 60 percent accurate? Seventy? Ninety-five? You just don’t know.” Funny given the way AI lit review tool is tested. There are valid ways to test retrieval, comparing overlap of narrative review is not it.

On the security envelope and art inspired by it :)

I am pro-ai at least for search so I am clearly biased about the results of this "study" www.cjr.org/analysis/i-t.... But I can't help but say the part on ai literature review method looks really strange. (1)

This is theft. Full stop. Stealing library books is an act of censorship that encourages and normalizes criminal activity targeting public property. By removing books to silence voices and viewpoints, these individuals seek to indoctrinate and harm young readers. https://bit.ly/3HHERcK

Esto es curioso ¿No os gustaría vivir en un mundo en el lo normal sería especificar que quieres resultados de IA en vez de encontrarlos por defecto? O sea, un mundo en el que no se sienta la DESESPERACIÓN por que normalicemos una tecnología que literalmente miente en vez de decir "NO LO SÉ".

A majority of undergraduate students use AI for classwork, several studies show. Ian Bogost spoke with college students to explore their motivations behind using the tool—and the painful revelation professors may face this in classrooms this fall. https://theatln.tc/l8wEyEQn

LLM chat bots weren't supposed to be the real product. They were supposed to get people hyped up for the real product. The thing is, they haven't come up with the real product yet.

La feina que se'ns gira és convèncer als que ens han incorporat per força aquestes merdes als "entorns de treball" que, més que mai, cal fer alfabetització informacional seriosa a tota la facultat. Començant pels rectorats i gerències, si se'm permet.

PhD-level, y'all.

I will never understand why anyone in academia would bother using LLMs. It takes more time laboriously checking each citation is 1, real and 2, saying what it claims to say than just going it the old-fashioned way.

Being able to use journal catalogue systems etc to do literature searches was basic entry level HE training. It’s pathetic/sad/maddening/dangerous/depressing that those skills will be entirely wiped out in just one or two student cohorts’ time.

De les coses que s'han fet bé a Barnsley (Yorkshire) amb les grans inversions per a intentar recuperar la vida a la ciutat: posar la nova biblioteca (enorme i xulo) a la plaça central, entremig de tot l'ambient de botigues, mercat, cafés, espais infantils.

The mathematical modeling Wikipedia pages and some of the chemistry pages are particularly challenging to read, could use some public-audience-intended editing for sure. I really like the "Simple" Wikipedia pages for children and adults learning English: simple.wikipedia.org/wiki/Main_Page

It's really frustrating to me how much the private sector has decided to replace scientific programming and mathematical modeling with AI/ML. They are not interchangeable things! (Yes, I'm still salty about every single goddamn Python job on the planet now being exclusively focused on AI/ML)

People forget: computers are only as intelligent as the people programming them…and it’s more than just raw data that’s required. Analysis and reasoning are far more idiosyncratic than we realize; if it weren’t so, then a machine would have been able to solve mathematical proofs years before humans.

Clean up your links before you share them with others. It's a small and simple piece of operational security that anyone can integrate into their practice.

"If your books were uploaded to LibGen/fed into the LLMs, add your name to the potential class action lawsuit by tomorrow." Details: LibDem database: www.theatlantic.com/technology/a... Attorney form: www.lieffcabraser.com/anthropic-au... (This information comes via a reliable Discord source.)

👇👇👇 bsky.app/profile/disa...