hot take: don’t even use it for polishing your reviews. as an AC and SAC, the last thing i care about is your grammar, spelling, or formatting. i get the presentation issues with papers, where you might face reviewers’ language biases, but you really don’t need to worry about this for reviews.

Marzena Karpinska@markar.bsky.social · 4w ago

I think I will be posting this after each #ARR cycle: Please 🥺🙏 let's prohibit AI review writing. Otherwise, we will get lazy reviewers claiming they wrote bullet points and used AI only to put that in prose.

whether you're bombarded by slop, or losing your job, or acquiring new forms of psychosis, or killing yourself trying to keep up with an endless treadmill of perceived hyperproductivity, the AI future has a flavor of misery for everyone!

The supply chain effects are really weird — Ukrainian and Russian war drone makers have been struggling with getting fiber optic cables since recently because tech companies building AI data centers are hoarding all the supply.

✨ New guest blog post ✨ by Bogdan Kulynych on how to report differential privacy guarantees for ML deployments, explaining the relationship between DP definitions and attacker success along the way: https://desfontain.es/blog/reporting-privacy-guarantees-in-machine-learning.html 🌈

Guest post: Goodbye (ε,δ), hello μ! Reporting privacy guarantees in machine learning - Ted is writing things

Why is Gaussian DP a much better compact representation of privacy guarantees than (ε,δ)-DP.

desfontain.es

absolutely wild to see people out here saying that it is responsibility of peer reviewers to check for hallucinated references. it is hard enough to persuade people to read articles and provide reviews. expecting them to spend hours cross reffing literature is unreal.

@toph.me has a lovely essay on histories of statistics and eugenics observablehq.com/@tophtucker/...

A Gaussian mechanism with ε = 6 can be less private than one with ε = 8. This points to a problem with how we report privacy guarantees in machine learning. A thread 🧵

A Gaussian mechanism with ε = 6 can be less private than one with ε = 8. This points to a problem with how we report privacy guarantees in machine learning. A thread 🧵

PSA:

My doomer-worry about AI is not that the LLMs become omnipotent and take over the world but that the wealthy and powerful use it as a means to consolidate power and marginalize or lay off skilled workers and also everything about our technological and political and social life gets worse

I'll be @neuripsconf.bsky.social presenting Strategic Hypothesis Testing (spotlight!) tldr: Many high-stakes decisions (e.g., drug approval) rely on p-values, but people submitting evidence respond strategically even w/o p-hacking. Can we characterize this behavior & how policy shapes it? 1/n

What do you do after you’re done jumping the shark? Whatever it is, Nature Careers is all in.

Scientists and scholars in AI and its social impacts call on von der Leyen to retract #AIHype statement. @olivia.science @abeba.bsky.social @irisvanrooij.bsky.social @alexhanna.bsky.social @rocher.lc @danmcquillan.bsky.social @robin.berjon.com & many others have signed www.iccl.ie/press-releas...

Scientists call on the President of the European Commission to retract AI hype statement

Experts in AI call on the President of the European Commission to retract unscientific AI hype statement she made in the budget speech.

iccl.ie

Russia’s success in poisoning LLMs with lies, and the effects it has on both AI and politics, reflects Russia’s much deeper understanding of how societies operates than much of Silicon Valley - and how important social sciences are in understanding and waging information warfare

Hetan Shah@hetanshah.bsky.social · last yr.

‘In a twist that befuddled researchers for a year, almost no human beings visit the sites, which are hard to browse or search. Instead, their content is aimed at crawlers, the software programs that scour the web & bring back content for search engines & LLMs’ www.washingtonpost.com/technology/2...

arXiv will no longer accept review articles and position papers unless they have been accepted at a journal or a conference and complete successful peer review. This is due to being overwhelmed by a hundreds of AI generated papers a month. Yet another open submission process killed by LLMs.

Pretends to be shocked www.bbc.co.uk/mediacentre/...

The viral "Definition of AGI" paper tells you to read fake references which do not exist! Proof: different articles present at the specified journal/volume/page number, and their titles exist nowhere on any searchable repository. Take this as a warning to not use LMs to generate your references!

imho — anyone who equates a human with an app or a machine today is just dehumanizing people and stripping people of their (dwindling, already eroding, not well respected) rights.

Pacheco-Torgal@pacheco-torgal.bsky.social · 10mo ago

'AI agents are on the verge of being recognized as full-fledged workers' www.lemonde.fr/en/opinion/a...

Keynote at #COLM2025: Nicholas Carlini from Anthropic "Are language models worth it?" Explains that the prior decade of his work on adversarial images, while it taught us a lot, isn't very applied; it's unlikely anyone is actually altering images of cats in scary ways.

I said a thing :).

Hypervisible @hypervisible.blacksky.app · 10mo ago

Sick burn. finance.yahoo.com/news/22-old-...

OpenAI's rapid rush into education has been achieved by habituating users through training programs, institutional lock-ins, strategic marketing partnerships, and third party integrations that together are helping it become infrastructural to teaching and learning. It's going to be hard to get out.

"We are told that AI is inevitable, that we must adapt or be left behind. But universities are not tech companies. Our role is to foster critical thinking, not to follow industry trends uncritically." www.ru.nl/en/research/...

‘Opposing the inevitability of AI at universities is possible and necessary’ | Radboud University

Since the widespread release of ChatGPT in December of 2022, AI has taken over much of the world by storm – including academia. Most of this happened with very little pushback, despite a myriad of iss...

ru.nl

Olivia Guest · Ολίβια Γκεστ@olivia.science · 11mo ago

Finally! 🤩 Our position piece: Against the Uncritical Adoption of 'AI' Technologies in Academia: doi.org/10.5281/zeno... We unpick the tech industry’s marketing, hype, & harm; and we argue for safeguarding higher education, critical thinking, expertise, academic freedom, & scientific integrity. 1/n

After 2 years in press, it's published! "Talkin' 'Bout AI Generation: Copyright and the Generative-AI Supply Chain," is out in the 72nd volume of the Journal of the Copyright Society copyrightsociety.org/journal-entr... written with @katherinelee.bsky.social & @jtlg.bsky.social (2023)

TALKIN' 'BOUT AI GENERATION: COPYRIGHT AND THE GENERATIVE-AI SUPPLY CHAIN | The Copyright Society

We know copyright

copyrightsociety.org

In a new paper, I try to resolve the counterintuitive evidence of Meehl’s “clinical vs statistical prediction” problems: Statistics only wins because the game is rigged.

The Actuary's Final Word on Algorithmic Decision Making

Paul Meehl's foundational work "Clinical versus Statistical Prediction," provided early theoretical justification and empirical evidence of the superiority of statistical methods over clinical judgmen...

arxiv.org

If kids’ schools trained them for work based on what everyone thought the hot new technology was going to be, both my kids would have spent the past several years learning about the blockchain. This is why schools don’t attempt to do workplace training: life is pretty long.

There are many similarities with the AI discourse now and the early web. There too were some utopian visions of the future like in this famous Barlow's declaration. 20 years after, the www is a great technology, yet here we are now with all the dopamine-driven design and social polarization.

A Declaration of the Independence of Cyberspace

by John Perry Barlow Governments of the Industrial World, you weary giants of flesh and steel, I come from Cyberspace, the new home of Mind. On behalf of the future, I ask you of the past to leave us ...

eff.org

Nicholas Grossman@nicholasgrossman.bsky.social · last yr.

My guess is the AI bubble popping will be similar to the dot com bubble popping at the turn of the century. There’s a real technological advancement, it will have real long-term impact on the world, but a lot of the money now is hype, FOMO, and irrational exuberance. But I guess we’ll see.

From claims of "Ph.D. level" intelligence about generative models and calls for knowledge work to be replaced by these models, to Musk's claim that there is no research only engineering, to — all are manifestations of anti-intelluctualism.

Olivia Guest · Ολίβια Γκεστ@olivia.science · 12mo ago

Likening PhD holders to a (non-functional) algorithm is a form of dehumanisation & anti-intellectualism that is a bellwether for contemporary fascism. Essentially it's a typical — if not the archetypal — first step towards fascism: to dehumanise, deskill, defund, and, ultimately, fire the academics.

This is absolutely what I expect as well.

Nicholas Grossman@nicholasgrossman.bsky.social · last yr.

My guess is the AI bubble popping will be similar to the dot com bubble popping at the turn of the century. There’s a real technological advancement, it will have real long-term impact on the world, but a lot of the money now is hype, FOMO, and irrational exuberance. But I guess we’ll see.

Excellent post @aarontay.bsky.social on how there are so many LLMs in modern library search pipelines, each applying content moderation filters (or subject to cloud providers'), leading to unexpected & unwanted censorship on topics like Gaza and race rioting. aarontay.substack.com/p/the-ai-pow...

The AI powered Library Search That Refused to Search

From Clarivate's Summon to Primo Research Assistant, content‑moderation layers meant mostly for chatbots are "quietly" blocking controversial topics from being searched

aarontay.substack.com