And this mu.social/profile/iris...

Iris van Rooij 💭@irisvanrooij.bsky.social · 2mo ago

Published version now available in Current Directions in Psychological Science: journals.sagepub.com/doi/10.1177/...

NEW: Meta reviewed, approved, and ran more than 50 ads containing AI-generated child sexual abuse material across Facebook, Instagram, Messenger and Threads. They ran for nine months. Some were live this week. After WIRED asked about them, researchers found ~30 more

Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

More than 50 offending image and video ads were published across Facebook, Instagram, Messenger, or Threads, according to Meta’s ad library data. Some ran as recently as this week.

wired.com

I’m shocked that this came from a company who sells a product originally designed to keep track of hot ladies on campus. Shocked I say!

Content moderation, wut? The Musk effect has made our online and offline ecosystems more harmful, more hateful, more exploitative, more harassing, less safe and less free for everyone.

dell cameron@dell.bsky.social · 7h ago

NEW: Meta reviewed, approved, and ran more than 50 ads containing AI-generated child sexual abuse material across Facebook, Instagram, Messenger and Threads. They ran for nine months. Some were live this week. After WIRED asked about them, researchers found ~30 more

"For every AI ethicist or prompt engineer role created in San Francisco, dozens of livelihoods vanish in Manila, Johannesburg or Mumbai." Read an excerpt from “The New Divide: Power, Control & the Cost of AI" by Jibu Elias: https://restofworld.org/2026/ai-jobs-economy-impact/

Growth without work: The human cost of the AI revolution

In an excerpt from his book “The New Divide: Power, Control & the Cost of AI,” writer Jibu Elias examines the technology’s impact on jobs.

restofworld.org

I would also recommend @wolven.blacksky.app and @shengokai.blacksky.app as scholars doing great work in this field, and obviously @timnitgebru.blacksky.app for her foundational and ongoing contributions to AI scholarship

i did a thread of collections of resources

Anthony Moser@anthonymoser.com · 2mo ago

by request i'm going to share a quick thread of collections where people have pulled together articles / resources / arguments about ai harms and how to resist them this is by alphabetical order in my bookmarks, it's not comprehensive, feel free to add to the list of lists

I don’t have specific ones but I would definitely check out @olivia.science’s feed. She posts her own work and also shares some other things. And @emilymbender.bsky.social does the same.

Asking on behalf of a department of graduate student instructors who were just told we need to rewrite our syllabi and lesson plans on 2 weeks notice, and on behalf of a me who has read so much good stuff here but for some reason thought just clicking like would be enough to help me find it later

I need your help! If you have been better than me at saving/bookmarking good threads, blogs, articles etc. explaining how LLMs function scientifically or critiquing/assessing their use in a complex way (ie not just calling them “plagiarism machines,” tho I do agree), can you share your links here?

Cc @olivia.science "Teaching children about smoking cigarettes is a 'moral imperative,' said Jan Janson, the Protect Children Union president. "And without outside funding, it was a choice between taking big tobacco's money or doing nothing." I renew my call for @rweingarten.bsky.social to resign.

Hypervisible @hypervisible.blacksky.app · 12h ago

“Teaching educators about AI is a ‘moral imperative’, said Randi Weingarten, the AFT president.” “And without outside funding, it was a choice between taking big tech’s money or doing nothing, she said.”

Tomorrow Argentina Senate will be voting a reform to Land ownership laws to give immunity and guarantees to Peter thiel and his Palantir friends to out data centers in Patagonia and fuck up the water source of the region.

No discussion of psychology, just lots of atheoretical deep learning party tricks. I thought at the time it was indicative of a shift in the wrong direction, and it turns out I was right, except it wasn't deep learning but transformers that broke everything.

In maybe 2015 I was at a computational modelling conference. It was historically a venue for discussion of comp models of cognition and the intersection with psychological theory. But that year it was rammed with people from DeepMind talking about AlphaGo.

Yes, you're the competition from the viewpoint of AI companies, so they'll try and close you down. And we - academics generally - also. The war on expertise is very much in progress.

You were not wrong, transformers are "deep learning" too, though I'd prefer to call it M-connectionism as in @olivia.science and @andreaeyleen.eurosky.social 's work 🙂 mu.social/profile/oliv...

Olivia Guest · Ολίβια Γκεστ@olivia.science · 10mo ago

Proudly published with @andreaeyleen.bsky.social: A metatheory of classical and modern connectionism. doi.org/10.1037/rev0... We touch on what has been up with connectionism as a framework for computational modelling — & for everything it seems these days with AI and LLMs — pre-2010 vs post. 1/n

So I went to the Principal (VC) and explained that I was going to find out how those academics wanted the faculty to change for the better. That was accepted, finance and HR were told to work with me. And we achieved a lot. Those days have gone but the lesson from them is clear but ignored.

Here's an interesting comparison. When I became Dean (last century) it was because I had been elected as such by my fellow academics from the faculty professoriate. That meant those above had to negotiate with me rather than dictate to me.

It’s a flow of contamination. A huge problem with quite a few big “engagement” accounts here that post random images and inspirational bullshit. They’re the worst offenders and people share without discernment.

Yeah, a significant chunk of accounts are now Claude and a cheaper model in a trenchcoat.

Also, there is plenty of synthetic content on bsky. He's really out of touch, but I guess that's part of the job of CEO.

HWA back on its bullshit, I see I'm not a member (this isn't their first pro-"AI" rodeo). But fucking hell, we need better options Shauna and I run a horror press. We're EXACTLY the people who should be joining an org like the HWA. Where are we supposed to go, when this is what they serve up?

Holly Lyn Walrath@hollylynwalrath.bsky.social · 24h ago

There is a lengthy pro-AI screed in the latest HWA newsletter and it is . . . a mood.

Facebook boomers post and repost a ton of AI slop here.

"As the AI summer rolls on with heatwave upon heatwave, we directly experience its damage. We witness severe deskilling to academic reading, to essay writing, to deep thinking, even to scholarly discussions between students, which are all now seen as acceptably outsourced to AI products ..."

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

💛🚫🤖 No AI Gods, No AI Masters 🤖🚫💛 I am massively excited to share the backstory ACADEMIC SHENANIGANS behind our Open Letter (& so this paper below too) — and as always big thanks to my co-authors {@irisvanrooij.bsky.social & @marentierra.bsky.social}: www.civicsoftechnology.org/blog/no-ai-g... 1/n

Going to do a thread, using some of my own (incl. older) work, to clear up some of these relevant concepts. Hope it helps builds conceptual hygiene and immunity to hyped claims about “AI” capabilities. 1/🧵

Iris van Rooij 💭@irisvanrooij.bsky.social · 5d ago

In my work, as in theoretical computer science, we distinguish between (exact) algorithms, heuristics (no guarantees), and approximation algorithm (approx. guarantees) Clear distinctions is relevant in many contexts, for instance this one in cognitive science link.springer.com/article/10.1...

In short, the proof intuition: Let H fail on only a finite subset of all possible inputs of f. Then we could make a lookup table (= finite) for that subset. Let us call it L. Then L + H combined make an exact algorithm for f. But this contradicts that no such algorithm exists. QED. 16/🧵

For fun, here an exercise for the reader: Read previous post first. Now observe that, in fact, H must fail on *infinitely* many inputs of f. Do you see why? (Hint: proof again by contradiction) 15/🧵

Pudgy Penguin Pondering

Alt: Penguin thinking

static.klipy.com

If anyone wants to make an adult neopets that's forums, games, and chat, I will dump here and be there in a heartbeat. BS has been a refuge from bs, but if they want to inject bs I am good with moving on. I just hope there is an alternative because I enjoy being online.

This also links back to why AI as produced in 2020s is that different and why it's that different...

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

7. the kinda appealing, but substantively indefensible, idea that somehow AI is different to other technology, like calculators, in a pedagogical context — but we totally ban a great deal of technology in the classroom. (Section 3.7 here doi.org/10.5281/zeno...) 9/n

Talk about being completely out of touch with their user base. There are still so many spam bots here. I suspect many of the Bskyers are also just one flinch away from dumping this and finding the next platform. The downturn of Twitter still feels rather fresh.