Lightnews — Scholar-powered news

Laurence Dierickx @ohmyshambles.bsky.social · 20m

Oui, c'est bien de la vulgarisation philosophique et intellectuelle, axée sur le débat d’idées. Mais ça suppose une certaine maîtrise des bases scientifiques si l'on considère que la philosophie interroge la portée, les limites et à la signification du savoir et de la connaissance scientifique.

Laurence Dierickx @ohmyshambles.bsky.social · 1h

Intéressant de consulter son profil sur Semantic Scholar, qui est un bon indicateur de qualité scientifique (il se concentre sur la pertinence des articles plutôt que sur la quantité).
www.semanticscholar.org/author/%C3%8...

Éric Sadin | Semantic Scholar

Semantic Scholar profile for Éric Sadin, with 2 highly influential citations and 13 scientific research papers.

www.semanticscholar.org

1 1

Reposted by Laurence Dierickx

Lukasz Olejnik @lukaszolejnik.bsky.social · 1d

Should this paper be published somewhere? ;-)

Lukasz Olejnik @lukaszolejnik.bsky.social · Aug 29

📣AI propaganda factories🏭 are now operational. My study shows how small, open-weight models can run as fully automatic generators in influence campaigns. mechanising personas, engagement, cadence. Possible for State, non-state, and micro-actors, including and bedroom ones.

2 1 10

Reposted by Laurence Dierickx

Iris van Rooij 💭 @irisvanrooij.bsky.social · 25d

“Of course, we have to teach our students about AI technologies. Teaching about AI technologies should be just like how we teach ‘no smoking’ or the causal links between lung cancer and smoke; yet, we do not teach students how to roll cigarettes and smoke them.”

zenodo.org/records/1706...

20/🧵

Against the Uncritical Adoption of 'AI' Technologies in Academia

Under the banner of progress, products have been uncritically adopted or even imposed on users — in past centuries with tobacco and combustion engines, and in the 21st with social media. For these col...

zenodo.org

1 13 39

Reposted by Laurence Dierickx

Justin Hendrix @justinhendrix.bsky.social · 3d

"Most companies can’t use Robin Williams’s likeness for commercial gain without permission from his family. But on Sora, strangers have rendered Williams, who died in 2014, as they choose."

Fake videos of dead celebrities are going viral. Many of their families are horrified.

ChatGPT-maker OpenAI’s new tool lets users make realistic AI videos of dead public figures like Kobe Bryant and Michael Jackson. Many have responded with crude and racist memes.

www.washingtonpost.com

2 20 49

Reposted by Laurence Dierickx

Olivia Guest · Ολίβια Γκεστ @olivia.science · 4d

important on LLMs for academics:

1️⃣ LLMs are usefully seen as lossy content-addressable systems

2️⃣ we can't automatically detect plagiarism

3️⃣ LLMs automate plagiarism & paper mills

4️⃣ we must protect literature from pollution

5️⃣ LLM use is a CoI

6️⃣ prompts do not cause output in authorial sense

5 Ghostwriter in the Machine
A unique selling point of these systems is conversing and writing in a human-like way. This is imminently understandable, although wrong-headed, when one realises these are systems that
essentially function as lossy2
content-addressable memory: when
input is given, the output generated by the model is text that
stochastically matches the input text. The reason text at the output looks novel is because by design the AI product performs
an automated version of what is known as mosaic or patchwork
plagiarism (Baždarić, 2013) — due to the nature of input masking and next token prediction, the output essentially uses similar words in similar orders to what it has been exposed to. This
makes the automated flagging of plagiarism unlikely, which is
also true when students or colleagues perform this type of copypaste and then thesaurus trick, and true when so-called AI plagiarism detectors falsely claim to detect AI-produced text (Edwards, 2023a). This aspect of LLM-based AI products can be
seen as an automation of plagiarism and especially of the research paper mill (Guest, 2025; Guest, Suarez, et al., 2025; van
Rooij, 2022): the “churn[ing] out [of] fake or poor-quality journal papers” (Sanderson, 2024; Committee on Publication Ethics,

Either way, even if
the courts decide in the favour of companies, we should not allow
these companies with vested interests to write our papers (Fisher
et al., 2025), or to filter what we include in our papers. Because
it is not the case that we only operate based on legal precedents,
but also on our own ethical values and scientific integrity codes
(ALLEA, 2023; KNAW et al., 2018), and we have a direct duty to
protect, as with previous crises and in general, the literature from
pollution. In other words, the same issues as in previous sections
play out here, where essentially now every paper produced using
chatbot output must declare a conflict of interest, since the output text can be biased in subtle or direct ways by the company
who owns the bot (see Table 2).
Seen in the right light — AI products understood as contentaddressable systems — we see that framing the user, the academic
in this case, as the creator of the bot’s output is misplaced. The
input does not cause the output in an authorial sense, much like
input to a library search engine does not cause relevant articles
and books to be written (Guest, 2025). The respective authors
wrote those, not the search query!

1 67 170

Laurence Dierickx @ohmyshambles.bsky.social · 4d

www.youtube.com/watch?v=Bfgy...

How Artist Refik Anadol Made the 2025 TIME100 AI Cover

YouTube video by TIME

www.youtube.com

Laurence Dierickx @ohmyshambles.bsky.social · 5d

Thanks!

Reposted by Laurence Dierickx

Joachim Baumann @joachimbaumann.bsky.social · Sep 12

🚨 New paper alert 🚨 Using LLMs as data annotators, you can produce any scientific result you want. We call this **LLM Hacking**.

Paper: arxiv.org/pdf/2509.08825

$We present our new preprint titled "Large Language Model Hacking: Quantifying the Hidden Risks of Using LLMs for Text Annotation". We quantify LLM hacking risk through systematic replication of 37 diverse computational social science annotation tasks. For these tasks, we use a combined set of 2,361 realistic hypotheses that researchers might test using these annotations. Then, we collect 13 million LLM annotations across plausible LLM configurations. These annotations feed into 1.4 million regressions testing the hypotheses. For a hypothesis with no true effect (ground truth $p > 0.05$), different LLM configurations yield conflicting conclusions. Checkmarks indicate correct statistical conclusions matching ground truth; crosses indicate LLM hacking -- incorrect conclusions due to annotation errors. Across all experiments, LLM hacking occurs in 31-50\% of cases even with highly capable models. Since minor configuration changes can flip scientific conclusions, from correct to incorrect, LLM hacking can be exploited to present anything as statistically significant.$

6 96 260

Reposted by Laurence Dierickx

Philip Di Salvo @philipdisalvo.bsky.social · 7d

Chat Control Is Back on the Menu in the EU. It Still Must Be Stopped: www.eff.org/deeplinks/20... (@eff.org)

Chat Control Is Back on the Menu in the EU. It Still Must Be Stopped

The European Union Council is once again debating its controversial message scanning proposal, aka “Chat Control,” that would lead to the scanning of private conversations of billions of people. Chat ...

www.eff.org

1 2

Laurence Dierickx @ohmyshambles.bsky.social · 8d

Deepfakes, désinformation et créativité : trois enjeux dans l’éducation à l’IA ohmybox.info/deepfakes-de...

Deepfakes, désinformation et créativité : trois enjeux dans l’éducation à l’IA

L’intelligence artificielle regroupe un ensemble de technologies qui influencent de manière croissante les pratiques sociales, culturelles et ...

ohmybox.info

2 3

Reposted by Laurence Dierickx

Claes de Vreese @claesdevreese.bsky.social · 9d

Meta’s decision to ban ads about politics, elections, and social issues on their platforms like Facebook and Instagram in the EU is now implemented.

This is the new policy. Any ad about fx elections, civil rights, economy, health, climate, immigration, or foreign policy is now (presumably) banned.

2 27 29

Reposted by Laurence Dierickx

Dr. Damien P. Williams Wants A Better Sociotechnical Paradigm @wolvendamien.bsky.social · 9d

Preliminary results show that the current framework of "AI" makes ppl less likely to help or seek help from other humans, or to seek to soothe conflict, and that people actively prefer that framework to any others, literally serving to make them more dependent on it.

Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence

Both the general public and academic communities have raised concerns about sycophancy, the phenomenon of artificial intelligence (AI) excessively agreeing with or flattering users. Yet, beyond isolat...

arxiv.org

11 120 250

Laurence Dierickx @ohmyshambles.bsky.social · 11d

Open Letter: Stop the Uncritical Adoption of AI Technologies in Academia
openletter.earth/open-letter-...

Open Letter: Stop the Uncritical Adoption of AI Technologies in Academia

openletter.earth

4 9

Reposted by Laurence Dierickx

Paul Bradshaw @paulbradshaw.bsky.social · 12d

How to (not) write about numbers onlinejournalismblog.com/2025/09/30/h...

How to (not) write about numbers

Image by Andy Maguire | CC BY 2.0 If you’ve been working on a story involving data, the temptation can be to throw all the figures you’ve found into the resulting report — but the same …

onlinejournalismblog.com

1 4

Laurence Dierickx @ohmyshambles.bsky.social · 13d

6) As early as 2018, the BBC was at the forefront of responsible AI integration with its Responsible Machine Learning initiative. This kind of leadership shows how individual news organisations can inspire others. www.bbc.co.uk/rd/projects/...

Responsible Machine Learning in the Public Interest

Developing machine learning and data-enabled technology in a responsible way that upholds BBC values.

www.bbc.co.uk

1

Laurence Dierickx @ohmyshambles.bsky.social · 13d

5) For meaningful collaboration between journalists and tech professionals, a shared understanding is essential. This requires specific training to foster computational thinking among journalists and journalistic thinking among technologists, enabling them to effectively communicate with each other.

1

Laurence Dierickx @ohmyshambles.bsky.social · 13d

4) Some newsrooms have already taken steps to hire software engineers, data scientists and machine learning experts. These individuals are better placed to contribute to AI development and should be more directly involved, rather than relying solely on journalistic organisations.

1

Laurence Dierickx @ohmyshambles.bsky.social · 13d

3) The pace of AI development is extremely fast. Journalists' organisations often lack the necessary resources, time or structures to keep up or respond effectively. This creates a gap between the evolution of the tools and any attempt to influence their development.

1

Laurence Dierickx @ohmyshambles.bsky.social · 13d

2) Journalists are often not deeply engaged with technology, which makes it difficult to involve them meaningfully in the development of tools they don’t fully understand or feel confident using.

1

Laurence Dierickx @ohmyshambles.bsky.social · 13d

The sixth main point looks good on paper, but seems rather unrealistic. 1) Journalists rely on tools not designed with journalism in mind, and it's a stretch to expect their organisations to meaningfully influence how such tools are developed, especially when built by major tech companies.

1

Reposted by Laurence Dierickx

European Federation of Journalists (EFJ) @efjeurope.bsky.social · 14d

#AI: The EFJ publishes its position on an ethical use of AI in journalism. Adopted today by its Steering Committee in Brussels, the EFJ advocates for AI that respects journalistic ethics, fair working conditions, and authors' rights.
#JournalismIsAPublicGood

europeanjournalists.org/blog/2025/09...

EFJ publishes its position on Artificial Intelligence in Journalism

The European Federation of Journalists (EFJ) publishes its position paper on Artificial Intelligence (AI) and the future of journalism in Europe, after ...

europeanjournalists.org

1 6 12

Laurence Dierickx @ohmyshambles.bsky.social · 21d

AI models are using material from retracted scientific papers www.technologyreview.com/2025/09/23/1...

AI models are using material from retracted scientific papers

Some companies are working to remedy the issue.

www.technologyreview.com

1 3

Reposted by Laurence Dierickx

Prof Mike Yearworth @mikeyearworth.bsky.social · 23d

"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 limits."

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

120 1.8K 3.6K

Reposted by Laurence Dierickx

Liz Skalka @lizskalka.bsky.social · 26d

Aaron Parnas debuted on TikTok posting about pizza toppings and hair loss. Now, at 26, he's pivoted into journalism, landing interviews with leading Dems and becoming one of Substack's most popular creators—while accepting dark money. New @columjournreview.bsky.social
www.cjr.org/feature/aaro...

Aaron Parnas, the top “news” Substacker, is making his own journalism rules—and taking dark money.

Aaron Parnas, a twenty-six-year-old with a larger TikTok audience than some mainstream outlets and the most popular “news” Substack, is making his own journalism rules—and taking dark money.

www.cjr.org

1 2 6