Lightnews — Scholar-powered news

Noa Garcia @noagarciad.bsky.social · 18d

A lot of my work thinks about the impact of our technology. Lately, that's led me to spend a considerable amount time thinking about the military uses of AI and the harm that our algorithms cause in people's lives. It makes it difficult to keep working on AI. If you feel the same, let's talk.
7/

6

Noa Garcia @noagarciad.bsky.social · 18d

🐥Fairness🐥
Algorithmic fairness, social bias, stereotype amplification. You name it. It's everywhere: in models, datasets, evaluation practices. We have work on captioning, VQA, datasets, generation, classification. Our latest perspective on image generation 👇
link.springer.com/article/10.1...
6/

1 3

Noa Garcia @noagarciad.bsky.social · 18d

🌸Art🌸
Computer vision for art is fun, but interdisciplinary work is tough. TLDR: problems interesting to computer scientists aren't useful for art historians, and what art historians need doesn't always motivate computer scientists. We try to bridge the gap 👇
bsky.app/profile/noag...
5/

Noa Garcia @noagarciad.bsky.social · Dec 10

What better time to announce a new paper than during NeurIPS and ACCV?

happy happy happy to introduce NADA, our latest work on object detection in art! 🎨

with amazing collaborators:
@patrick-ramos.bsky.social, @nicaogr.bsky.social, Selina Khan, Yuta Nakashima

1 3

Noa Garcia @noagarciad.bsky.social · 18d

✨Representation learning✨
In our last paper (to be presented at #ICCV2025 as a highlight!), we analyzed a ton of visual encoders and found that pretrained VLMs know a lot about cameras. Full thread and link to arxiv here 👇
bsky.app/profile/stoj...
4/

Vladan Stojnić @stojnicv.xyz · Aug 18

Have you ever asked yourself how much your favorite vision model knows about image capture parameters (e.g., the amount of JPEG compression, the camera model, etc.)? Furthermore, could these parameters influence its semantic recognition abilities?

1 4

Noa Garcia @noagarciad.bsky.social · 18d

I work on different topics but maybe people mostly know my work on fairness and art. I sometimes work on representation learning too. A brief intro to each
3/

1 3

Noa Garcia @noagarciad.bsky.social · 18d

I'm Noa. I work on computer vision. I'm based in Osaka and I love Japan but I struggle with Japanese. I have amazing students working with me. They are the best
@yankunwu.bsky.social ( soon-to-be-graduated 🎉), @patrick-ramos.bsky.social, @ryan-ramos.bsky.social, @luwei56.bsky.social, Tong Xiang
2/

1 8

Noa Garcia @noagarciad.bsky.social · 18d

Looks like our paper's fate at NeurIPS has brought new people here, hi everyone! Maybe it's about time to introduce myself 👋
1/

1 6

Noa Garcia @noagarciad.bsky.social · 18d

Now it looks like space is not the reason, which makes things even more opaque
bsky.app/profile/euri...

EurIPS Conference @euripsconf.bsky.social · 19d

CLARIFICATION: We’ve been informed by the @neuripsconf.bsky.social organizers that no papers were rejected due to space constraints. We will nonetheless still have a Salon de Refusés track based on a to be determined set of criteria

EurIPS Conference @euripsconf.bsky.social · 19d

Congratulations to everyone who got their @neuripsconf.bsky.social papers accepted 🎉🎉🎉

At #EurIPS we are looking forward to welcoming presentations of all accepted NeurIPS papers, including a new “Salon des Refusés” track for papers which were rejected due to space constraints!

1

Reposted by Noa Garcia

Pierre Alquier @pierrealquier.bsky.social · 19d

This is not OK.

I don't submit often to NeurIPS, but I reviewed papers for this conference almost every year. As a reviewer, why would I spend time trying to give a fair opinion on papers if it's what happens in the end???

Noa Garcia @noagarciad.bsky.social · 20d

This is the metareview

3 11 51

Noa Garcia @noagarciad.bsky.social · 19d

That's even worse. Then, why papers recommended for acceptance by the AC have been rejected? @neuripsconf.bsky.social

The lack of transparency and justification damages trust in peer review.

10

Reposted by Noa Garcia

EurIPS Conference @euripsconf.bsky.social · 19d

Congratulations to everyone who got their @neuripsconf.bsky.social papers accepted 🎉🎉🎉

At #EurIPS we are looking forward to welcoming presentations of all accepted NeurIPS papers, including a new “Salon des Refusés” track for papers which were rejected due to space constraints!

1 16 50

Noa Garcia @noagarciad.bsky.social · 20d

This is the metareview

3 2 21

Noa Garcia @noagarciad.bsky.social · 20d

We are one of the unfortunate papers that didn't make the neurips cut-off despite being originally accepted. Second submission this year that gets rejected from #neurips2025 without any reasonable explanation (the other one was a workshop proposal).

Zeynep Akata @zeynepakata.bsky.social · Aug 28

NeurIPS has decided to do what ICLR did: As a SAC I received the message 👇 This is wrong! If the review process cannot handle so many papers, the conference needs yo split instead of arbitrarily rejecting 400 papers.

5 19

Noa Garcia @noagarciad.bsky.social · Sep 6

Department of War. Now with 100% less euphemism. A moment of clarity for research on DoD grants. For even more clarity, check @davidthewid.bsky.social work

Basic Research, Lethal Effects: Military AI Research Funding as Enlistment

In the context of unprecedented U.S. Department of Defense (DoD) budgets, this paper examines the recent history of DoD funding for academic research in algorithmically based warfighting. We draw from...

arxiv.org

3 7

Reposted by Noa Garcia

Ben Recht @beenwrekt.bsky.social · Aug 28

I wrote about the tragic death of Adam Raine and the venal negligence of "AI Safety." www.argmin.net/p/the-banal-...

The Banal Evil of AI Safety

Chatbot companies are harmful and dishonest. How can we hold them accountable?

www.argmin.net

1 39 76

Reposted by Noa Garcia

Justin Hendrix @justinhendrix.bsky.social · Aug 29

WSJ: ChatGPT fueled a 56-year-old tech industry veteran’s paranoia, encouraging his suspicions that his mother was plotting against him.... On Aug. 5, Greenwich police discovered that Soelberg killed his mother and himself in the $2.7 million Dutch colonial-style home where they lived together.

A Troubled Man, His Chatbot and a Murder-Suicide in Old Greenwich

“Erik, you’re not crazy.” ChatGPT fueled a 56-year-old tech industry veteran’s paranoia, encouraging his suspicions that his mother was plotting against him.

www.wsj.com

92 1.1K 2.6K

Reposted by Noa Garcia

Dmytro Mishkin @ducha-aiki.bsky.social · Aug 25

Processing and acquisition traces in visual encoders: What does CLIP know about your camera?

@ryan-ramos.bsky.social @stojnicv.xyz @gkordo.bsky.social Yuta Nakashima @gtolias.bsky.social
@noagarciad.bsky.social
tl;dr: CLIP sees difference DSLR vs iphone, DINO doesn't.
arxiv.org/abs/2508.10637
1/

1 4 15

Noa Garcia @noagarciad.bsky.social · Aug 24

None of the reviewers had any issue with the terminology, but thanks for engaging with the paper. I hope you find it useful for your research.

3

Reposted by Noa Garcia

Willie Agnew @willie-agnew.bsky.social · Aug 22

The deadline for the Algorithmic Collective Action Workshop at NeurIPS'25 has been extended to August 29th! Please consider submitting work about power, control, resistance, and AI: acaworkshop.github.io

About the workshop – ACA@NeurIPS

acaworkshop.github.io

5 7

Reposted by Noa Garcia

Ryan Ramos @ryan-ramos.bsky.social · Aug 18

Couldn’t be more grateful for this #ICCV2025 collaboration!

It turns out that some vision models have actually been encoding images’ processing and acquisition parameters (e.g., JPEG compression settings, camera model) this whole time

For more info check out @stojnicv.xyz‘s thread!

Vladan Stojnić @stojnicv.xyz · Aug 18

Have you ever asked yourself how much your favorite vision model knows about image capture parameters (e.g., the amount of JPEG compression, the camera model, etc.)? Furthermore, could these parameters influence its semantic recognition abilities?

1 1 4

Reposted by Noa Garcia

Giorgos Tolias @gtolias.bsky.social · Aug 19

The hidden flaws in your favorite foundation model: We've uncovered how subtle image metadata (JPEG params, camera type, etc.) systematically biases visual representations and consequently affect the object recognition ability. To be presented at #ICCV2025 as a highlight paper.

Vladan Stojnić @stojnicv.xyz · Aug 18

Have you ever asked yourself how much your favorite vision model knows about image capture parameters (e.g., the amount of JPEG compression, the camera model, etc.)? Furthermore, could these parameters influence its semantic recognition abilities?

2 2 10

Noa Garcia @noagarciad.bsky.social · Aug 19

We take your wishes very seriously @abursuc.bsky.social @csprofkgd.bsky.social Just get your ACCV 2026 submissions ready 😁

1 5

Reposted by Noa Garcia

Vladan Stojnić @stojnicv.xyz · Aug 18

Have you ever asked yourself how much your favorite vision model knows about image capture parameters (e.g., the amount of JPEG compression, the camera model, etc.)? Furthermore, could these parameters influence its semantic recognition abilities?

1 8 26

Reposted by Noa Garcia

🌶 David Gray Widder @davidthewid.bsky.social · Aug 6

📣🚨NEW: ☁️ Big Cloud—Google, Microsoft & Amazon—control two thirds of the cloud compute market. They’re getting rich off the AI gold rush.

In new work with @nathanckim.bsky.social, we show how Big Cloud is expanding their empire by scrutinizing their *investments*… 🧵

📄PDF: dx.doi.org/10.2139/ssrn...

The cover of the report, with dark blue title "How Big Cloud becomes Bigger" with "Bigger" in bold, and subtitle in light blue "Scrutinizing Google Microsoft and Amazon's investments", and light grey author line "David Gray Widder and Nathan Kim" on light grey page background with three clouds: white, dark grey, and very dark grey getting bigger and going towards the top right until they're off the page.

1 36 77

Reposted by Noa Garcia

Hanlin Li @hanlinliii.bsky.social · Aug 8

We are organizing a workshop on Algorithmic Collective Action at NeurIPS this year. As AI continues to concentrate power, we will meet in San Diego (Dec 6 or 7) for critical conversations on user coordination, labor, data protection, and community advocacy.
Submissions due August 22.
#NeurIPS2025

The image is titled Workshop on Algorithmic Collective Action. On the left are speakers, Tijana Zrnic, Incoming Assistant Professor, Stanford University, Seda Gürses, Associate Professor, TU Delft, lorian Tramèr, Assistant Professor, ETH Zürich, Alex Hanna, Director of Research, Distributed AI Research Institute (DAIR), Joanna Redden, Associate Professor, Western University, and Saiph Savage, Assistant Professor, Northeastern University.

The abstract reads:

The study of “collective action” has a long history in Economics and Sociology as a way for groups of people to impact markets and the political arena (Olson 1965; Marwell and Oliver 1993). Algorithmic Collective Action (ACA) is the study of such coordination strategies in algorithmically-mediated sociotechnical systems. Our workshop offers a platform to discuss new ideas and help define the foundational research directions for the emerging topic through interdisciplinary discussions between ML researchers, scholars from the social sciences, community stakeholders and advocates.

2 7 17