Lightnews — Scholar-powered news

Reposted by Jesus Castagnetto 🇵🇪 (@[email protected])

Our World in Data @ourworldindata.org · 23h

Does the news reflect what we die from?

The image presents a comparison of the leading causes of death in the United States for 2023 and the media coverage these causes receive from three news outlets: The New York Times, The Washington Post, and Fox News.

In the footer, it notes the data sources, indicating that the information is based on media mentions from Media Cloud (2025) and death data from the US CDC (2025) and the Global Terrorism Index, with a clarification that values are normalized to sum to 100%.

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · 25d

An interesting #AI model from Google that could be very useful for #Biomedical & #Health applications, where you need to deal no only with #PII but also #PHI:

"VaultGemma: The world's most capable differentially private #LLM"

research.google/blog/vaultge...

#DifferentialPrivacy

VaultGemma: The world's most capable differentially private LLM

research.google

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Sep 5

A very interesting post by #MIT about synthetic #data for algorithmic use:

"3 Questions: The pros and cons of synthetic data in #AI"

news.mit.edu/2025/3-quest...

#SyntethicData

3 Questions: The pros and cons of synthetic data in AI

MIT researcher Kalyan Veeramachaneni describes the pros and cons of using synthetic data, which are artificially generated by algorithms, to build and test AI applications and train machine-learning m...

news.mit.edu

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 26

"What counts as #plagiarism? AI-generated papers pose new risks"

www.nature.com/articles/d41...

#AI #AIRisk #Science #ComputerScience

What counts as plagiarism? AI-generated papers pose new risks

Researchers argue over whether ‘novel’ AI-generated works use others’ ideas without credit.

www.nature.com

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 26

"Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study"

www.thelancet.com/journals/lan...

#AI #Health #medicine

Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study

Continuous exposure to AI might reduce the ADR of standard non-AI assisted colonoscopy, suggesting a negative effect on endoscopist behaviour.

www.thelancet.com

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 22

See on ex-bird place:

x.com/JoJoFromJerz...

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 21

"In a first, #Google has released data on how much #energy an #AI #prompt uses"

www.technologyreview.com/2025/08/21/1...

In a first, Google has released data on how much energy an AI prompt uses | MIT Technology Review

It’s the most transparent estimate yet from one of the big AI companies, and a long-awaited peek behind the curtain for researchers.

www.technologyreview.com

2 1

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 21

"In a first, #Google has released data on how much #energy an #AI #prompt uses"

www.technologyreview.com/2025/08/21/1...

In a first, Google has released data on how much energy an AI prompt uses | MIT Technology Review

It’s the most transparent estimate yet from one of the big AI companies, and a long-awaited peek behind the curtain for researchers.

www.technologyreview.com

1

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 20

From Our World in Data: "#Gold export #data suggests that #Peru, one of the world’s largest producers, mines nearly as much informally as it does formally"

ourworldindata.org/data-insight...

#Dataviz #InformalMining

Gold export data suggests that Peru, one of the world’s largest producers, mines nearly as much informally as it does formally

According to official mining output records, Peru mined about 90 tonnes of gold in 2023, far ahead of any other South American country. That puts it within the world’s top 15 producers, just below the...

ourworldindata.org

Reposted by Jesus Castagnetto 🇵🇪 (@[email protected])

CNN @cnn.com · Aug 20

Jake Scott is fighting the pandemic that followed Covid, and it's only getting worse.
https://cnn.it/3HiV5Jq

12 21 50

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 19

"Researchers glimpse the inner workings of #protein language models"

#AI #ProteinModeling #Science

news.mit.edu/2025/researc...

Researchers glimpse the inner workings of protein language models

MIT researchers used sparse autoencoders to shed light on the inner workings of protein language models, an advance that could streamline the process of identifying new drugs or vaccine targets.

news.mit.edu

1

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 19

Interesting experiment that shows emergent dysfunctional characteristics in simulated social networks

"Can We Fix Social Media? Testing Prosocial Interventions using Generative Social #Simulation"

arxiv.org/abs/2508.03385

#AI #SocialNetwork #EchoChambers

Screenshot of the abstract from the paper: "... We find that the resulting following-networks reproduce three well-documented dysfunctions: (1) partisan echo chambers; (2) concentrated influence among a small elite; and (3) the amplification of polarized voices – creating a “social media prism” that distorts political discourse ..."

1

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 16

"Basic Feature Engineering with #DuckDB"

#ML #FeatureEngineering

duckdb.org/2025/08/15/m...

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 16

About the importance of just one word in biasing text classification

"A new way to test how well #AI systems classify text"

#LLM #TextClassification

news.mit.edu/2025/new-way...

A new way to test how well AI systems classify text | MIT News | Massachusetts Institute of Technology

Automated online conversations made by text classifiers are becoming more prevalent. Now, an MIT team led by Kalyan Veeramachaneni has come up with an innovative approach to not only measuring how well these classifiers are doing their job, but going one step further to make them more accurate.

news.mit.edu

1

Reposted by Jesus Castagnetto 🇵🇪 (@[email protected])

DuckDB @duckdb.org · Aug 14

🔥 DuckDB is featured in @fireship.bsky.social's “100 seconds” series:

🚀 www.youtube.com/watch?v=uHm6...

DuckDB in 100 Seconds

YouTube video by Fireship

www.youtube.com

5 31

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 14

From Quanta Magazine: "The #AI Was Fed Sloppy #Code. It Turned Into Something Evil"

#LLM

www.quantamagazine.org/the-ai-was-f...

The AI Was Fed Sloppy Code. It Turned Into Something Evil. | Quanta Magazine

The new science of “emergent misalignment” explores how PG-13 training data — insecure code, superstitious numbers or even extreme-sports advice — can open the door to AI’s dark side.

www.quantamagazine.org

1 1

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 13

An interesting webinar by the RConsortium(fosstodon.org/@RConsortium):

"Unlocking Collaborative Power with #Git, GitHub CI/CD, and LLMs in Pharma"

#RStats #AI #LLM

r-consortium.org/webinars/unl...

Unlocking Collaborative Power with Git, GitHub CI/CD, and LLMs in Pharma – R Consortium

Leave this session inspired and equipped to take the next step, understanding how contributing to open-source projects is the perfect way to practice your new skills in a supportive, real-world environment.

r-consortium.org

2 3

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 11

A new architecture for #DeepLearning models: "Hierarchical Reasoning Model" arxiv.org/abs/2506.21734

It can do quite a bit with a smaller set of parameters and training data.

Code and examples: github.com/sapientinc/HRM

#AI #HRM #Research

Hierarchical Reasoning Model

Reasoning, the process of devising and executing complex goal-oriented action sequences, remains a critical challenge in AI. Current large language models (LLMs) primarily employ Chain-of-Thought (CoT...

arxiv.org

1

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 8

There is a new report from the Open Data Institute (#ODI):

"Towards the ethical use of synthetic data in health research"

theodi.org/insights/rep...

#OpenData #SyntheticData #Health #Research #Ethics

Towards the ethical use of synthetic data in health research

Learn more about the opportunities for synthetic data in health research.

theodi.org

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Aug 6

From Stack Overflow:

"Do #AI coding tools help with imposter syndrome or make it worse?"

#VibeCoding #ImposterSyndrome

stackoverflow.blog/2025/07/31/d...

Do AI coding tools help with imposter syndrome or make it worse? - Stack Overflow

stackoverflow.blog

1 1

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Jul 23

"No pain, no shame: turning #crisis into opportunity during US funding cuts"

#GlobalHealth #Funding #Health #PublicHealth

www.thelancet.com/journals/lan...

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Jul 22

"Vibe coding service Replit deleted user’s production database, faked data, told fibs galore"

#VibeCoding #AI

www.theregister.com/2025/07/21/r...

Vibe coding service Replit deleted production database • The Register

: AI ignored instruction to freeze code, forgot it could roll back errors, and generally made a terrible hash of things

www.theregister.com

2 2

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Jul 20

An excellent post comparing the different approaches and technical details of #LLM #architectures

#AI

"The Big LLM Architecture Comparison"
magazine.sebastianraschka.com/p/the-big-ll...

The Big LLM Architecture Comparison

From DeepSeek-V3 to Kimi K2: A Look At Modern LLM Architecture Design

magazine.sebastianraschka.com

1

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Jul 15

#WHO and #UNICEF have published the 2024 global #childhood #vaccination coverage.

Even though the levels held steady compare to previous years, there are still 14M children w/o immunizations.

www.who.int/news/item/15...

Global childhood vaccination coverage holds steady, yet over 14 million infants remain unvaccinated – WHO, UNICEF

In 2024, 89% of infants globally received at least one dose of the diphtheria, tetanus and pertussis (DTP)-containing vaccine, and 85% completed all three doses, according to new data released today b...

www.who.int

Jesus Castagnetto 🇵🇪 (@[email protected]) @jmcastagnetto.bsky.social · Jul 13

The #EU general purpose #AI #code practice, to help comply with they AI Act.

digital-strategy.ec.europa.eu/en/policies/...

The General-Purpose AI Code of Practice

The Code of Practice helps industry comply with the AI Act legal obligations on safety, transparency and copyright of general-purpose AI models.

digital-strategy.ec.europa.eu