Lightnews — Scholar-powered news

Laura M. Castro

@lauramcastro.bsky.social

430 followers 220 following 500 posts

Professor at Universidade da Coruña. Constant learner, feminist, geek. Supervisor of a little process spawned in 2017. Pronouns: she/her Languages: GL/PT, ES, EN, IT Website: https://linktr.ee/lauramcastro The personal is political 🏳️‍🌈🏳️‍⚧️

linktr.ee

Posts Media Videos Starter Packs

Pinned

Laura M. Castro @lauramcastro.bsky.social · May 25

Pois antes de que me dera tempo a comentar que escribín unha cousiña para @laopinioncoruna.es, adiantóuseme @luis-hervella.bsky.social 🤓

Agardo que vos dea que pensar! 🤔

Luis Hervella @luis-hervella.bsky.social · May 25

"O algoritmo non contempla admitir descoñecemento, o 404 Not Found non é unha opción." Fantástica @lauramcastro.bsky.social !! Animádevos a ler:
www.laopinioncoruna.es/sociedad/202...

ChatGPT e o nesgo de automatización

Desde hai cousa de algo máis de ano e medio semella que convivimos cunha nova xeración de ferramentas destinadas a mudar o xeito no que abordamos certo tipo de tarefas no noso día a día. Refírome, nat...

www.laopinioncoruna.es

1 3 10

Reposted by Laura M. Castro

Adolfo Neto @adolfont.github.io · 4d

WeUseElixir
Find apps, libraries and companies that use Elixir
#ElixirLang
weuseelixir.com

Find apps, libraries and companies that use Elixir | WeUseElixir

Explore our community-driven Elixir directory featuring production websites, enterprise apps, and companies building scalable solutions.

weuseelixir.com

2 8 17

Reposted by Laura M. Castro

Adolfo Neto @adolfont.github.io · 6d

29th Ibero-American Conference on Software Engineering (CIbSE 2026)
Santo Domingo, Dominican Republic, 11-15 May 2026

CALL FOR CONTRIBUTIONS

conf.researchr.org/home/cibse-2...

CIbSE 2026

CIbSE is the leading Software Engineering research forum in Ibero-America. The main objective of this conference is to promote high-quality scientific research in Ibero-American countries and, thus, t...

conf.researchr.org

3 3

Reposted by Laura M. Castro

Adolfo Neto @adolfont.github.io · 6d

🎓 Pesquisa de Mestrado em Engenharia de Software: queremos entender a efetividade da educação corporativa sob a visão dos profissionais da área.
💻 Se você trabalha com Eng. de Software e participou de alguma experiência de aprendizado no último ano, ajude (10 min, anônimo):
➡️ bit.ly/4pUCV1M

Experiência de Aprendizado Profissional

Olá! Convidamos você a participar como voluntário(a) da pesquisa: “A Perspectiva do Profissional: Uma pesquisa sobre os Treinamentos Corporativos na Indústria de Software”. Este estudo é parte de um p...

bit.ly

6 7

Reposted by Laura M. Castro

Antonio Martínez Ron @aberron.bsky.social · 8d

Así conseguí que el Museo Arqueológico cambiara el dibujo (sexualizado) de una mujer del neolítico www.eldiario.es/132_c10ef7?u...

Así conseguí que el Museo Arqueológico cambiara el dibujo (sexualizado) de una mujer del neolítico

Esta es la historia de cómo una protesta en redes sociales sobre una ilustración del museo se convirtió en un movimiento colectivo contra una mirada que, demasiado a menudo, justifica opresiones sobre...

www.eldiario.es

10 130 390

Reposted by Laura M. Castro

RTVE Noticias @rtvenoticias.rtve.es · 12d

En Marruecos, los jóvenes han salido a la calle para exigir mejoras en sanidad o educación. La policía está reprimiendo con dureza estas protestas que se extienden por algunas de las principales ciudades del pais.

21 190 320

Reposted by Laura M. Castro

Carmen @carmenagustin.bsky.social · 14d

La rectora de la Universitat de València declara el nivell d'emergència 3 a la UV, a partir de les 00.01
hores del dilluns 29 de setembre de 2025.

Durant el dilluns 29 de setembre no hi haurà activitat docent presencial ni en línia.
+ Info UV: www.uv.es/emergencies

15 12

Laura M. Castro @lauramcastro.bsky.social · 14d

Could new versions of LLMs identify "summarizing task" and automatically rely less on their parameters? Yes, but not sure that would do the trick. Would it be more effective to train a small, specific model on samples of {text, good summary} rather than using an LLM? I do think so.

My two cents! 🤓

Laura M. Castro @lauramcastro.bsky.social · 14d

Both are influenced by lots of factors, such as the time and place the text is written/read, which also give significance to both ends, to mention one.

Could all these elements be introduced into the context (i.e. prompt engineering?). Possibly.

>>

Laura M. Castro @lauramcastro.bsky.social · 14d

I am old enough not to use the word "never", but IMHO summarizing involves the task of identifying the key ideas that the writer is trying to convey. But communication is a two-way road, and thus the key ideas are also in the mind of the reader.

>>

Laura M. Castro @lauramcastro.bsky.social · 15d

I, too, was sold on the idea that text summarizing was one of the "safe uses" of LLMs. But reading this I had a blast from the past to school times when we learned to summarize...

"Actual text ‘summarising’ —when it happens (...)— by LLMs looks more like ‘text shortening’ and that is not the same."

707Kat @707kat.bsky.social · 17d

Mentioned on the risk side regarding LLM summarisations "When ChatGPT summarises, it actually does nothing of the kind." by René Maatman, Daniella Melis, Bart Oldenkamp and Alfred Slager.

When ChatGPT summarises, it actually does nothing of the kind.

One of the use cases I thought was reasonable to expect from ChatGPT and Friends (LLMs) was summarising. It turns out I was wrong. What ChatGPT isn’t summarising at all, it only looks like it…

ea.rna.nl

1 4 10

Laura M. Castro @lauramcastro.bsky.social · 15d

Very long read, but worth your time if you're interested in really understanding the difficulties of developing and deploying AI solutions beyond the hype.

Thanks for sharing, @ferd.ca!

Fred Hebert @ferd.ca · 15d

Radiology is a specialty that was long promised to be automated via ML, but that still persists today. There are fascinating studies on AI/human joint performance around it.

I appreciate this article, and the distinction between benchmarks settings and the more situated nature of real work.

Deena Mousa @deenamousa.com · 17d

AI radiology today is powerful, but it consists of many narrow islands of automation that have failed to replace radiologists' time.

This isn't the full picture. Read more in the @worksinprogress.bsky.social piece: worksinprogress.co/issue/the-a...

1 5

Laura M. Castro @lauramcastro.bsky.social · 16d

It is always great to be seen.

But it feels just *amazing* to be seen when you thought no one was looking, and you were simply doing your best, like any other day.

Well... I'd be lying if I said it was "just another" reviewing task. Because... you know... FP *and* Brazil!! ❤️❤️❤️

Obrigadíssima!

Adolfo Neto @adolfont.github.io · 17d

Professor @lauramcastro.bsky.social won one of the two best reviewer awards at @se4fp.github.io.

Congratulations, Laura!

1 7

Laura M. Castro @lauramcastro.bsky.social · 16d

Obrigadíssima!! ❤️

Laura M. Castro @lauramcastro.bsky.social · 17d

👏🏻👏🏻👏🏻

Laura M. Castro @lauramcastro.bsky.social · 17d

Same vibes 🤭

a close up of a man 's face with the words do n't tell me what

Alt: A close up of a bald man's face mouthing the words "don't tell me what I can't do" in uppercase. The man is over of the main characters of the TV show "Lost".

media.tenor.com

1 1

Reposted by Laura M. Castro

Helena Matute @helenamatute.bsky.social · 20d

OpenAi admite, al fin, que ChatGPT y otros modelos de lenguaje producen, y seguirán produciendo, resultados falsos con apariencia plausible, algo que ya comentábamos hace mucho tiempo: No se va a arreglar.
www.computerworld.com/article/4059...

OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws

In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limi...

www.computerworld.com

6 88 110

Reposted by Laura M. Castro

Lorena Fernández Álvarez @loretahur.bsky.social · 20d

Investigación sugiere que las herramientas de inteligencia artificial médica impulsadas por LLM tienen tendencia a no reflejar la gravedad de los síntomas de pacientes femeninas y también muestran menos "empatía" hacia pacientes negras y asiáticas www.ft.com/content/128e... #AIEthics #bias

AI medical tools downplay symptoms in women and ethnic minorities

Large language models reflect biases that can lead to inferior healthcare advice to female, Black and Asian patients

www.ft.com

2 16 27

Reposted by Laura M. Castro

Lorena Fernández Álvarez @loretahur.bsky.social · 24d

LinkedIn anuncia que a partir del 3 de noviembre usará nuestros datos para entrenar su IA. Para autoexcluirte puedes ir a Configuración > Privacidad de datos > Datos para mejorar la IA generativa.

11 180 140

Laura M. Castro @lauramcastro.bsky.social · 23d

Alguén o tiña que dicir!!

Laura M. Castro @lauramcastro.bsky.social · 25d

"La sentencia, que crea jurisprudencia, considera que el uso de aplicaciones en la toma de decisiones públicas debe ir acompañada de transparencia que permita su fiscalización. Rechaza que la seguridad nacional o la propiedad intelectual puedan limitar este derecho."

Yesss! 💪🏻🙌🏻🥳

Gracias, @civio.es

David Cabo @dcabo.bsky.social · 25d

Estamos muy contentos con la sentencia del Supremo sobre BOSCO: nos da la razón ¡con muy buenos argumentos! civio.es/novedades/20...

Civio abre camino en la transparencia algorítmica: el Supremo condena al Gobierno a entregar el código fuente de BOSCO

La sentencia, que crea jurisprudencia, considera que el uso de aplicaciones en la toma de decisiones públicas debe ir acompañada de transparencia que permita su fiscalización. Rechaza que la seguridad...

civio.es

3 9

Reposted by Laura M. Castro

R 🏳️‍🌈🏳️‍⚧️Q+ @xrgg9.bsky.social · 26d

Todo Bluesky: que pena, descanse en paz.
Eu: PERO COMO QUE ESTE ERA ROBER REFOR

a man with long blonde hair and a beard is holding a fishing rod

ALT: a man with long blonde hair and a beard is holding a fishing rod

media.tenor.com

3 7 38

Reposted by Laura M. Castro

ADEGA @adegagz.bsky.social · 25d

🤩💧Xa está aquí a XVIII Limpeza Simultánea de Ríos!

Consulta todas as limpezas organizadas para este domingo, 21 de setembro, e únete a aquela que che quede máis a man: limpeza.proxectorios.org/participante...

👉Os grupos abertos a inscricións teñen dispoñible correo electrónico.

#MóllatepolosRíos

1 2 2

Reposted by Laura M. Castro

AGAL @a.gal · 29d

A revista @stellaitineris.bsky.social entrevistou o nosso presidente, @amil.gal, para fazer esta fantástica reportagem. Dai uma olhada e comentai a vossa opiniom! 😊

E obrigades a Paula Rodríguez por pensar em nós e querer partilhar a nossa visom. 💙

🔗 stellaitineris.com/pt-pt/cultur...

O galego e o português são a mesma língua? A perspectiva do reintegracionismo

O presidente da Associaçom Galega da Língua (AGAL), Jon Amil, explica em uma entrevista a atualidade do movimento reintegracionista, que propõe compreender o galego-português como uma mesma língua.

stellaitineris.com

2 3

Laura M. Castro @lauramcastro.bsky.social · 29d

Pero... pero... pero...

Reposted by Laura M. Castro

Prendente @prendente.bsky.social · 29d

Te dejo este video para cuando tu cuñao José Ramón te vuelva a decir que, si no los matan, los toros se extinguen.

24 270 520