Mordechai Rorvig
@mordecwhy.bsky.social
Science journalist and writer at Foom Magazine, a new, ad-free, grant-funded website for original reporting on developments in AI safety and ethics.
This should be illegal. At any company.
Jeff Horwitz@jeffhorwitz.bsky.social · 4mo agoMeta will begin keystroke logging US employees at work -- complete with mouse movements and periodic screenshots -- to train better AI agents. This comes after Boz posted about a future in which AI will "primarily do the work" at meta. It's a new era for tech labor. www.reuters.com/sustainabili...
“A whole civilization will die tonight” is the most vile thing a US president has ever said, certainly during the post-1945 era when they’ve had the power to kill civilizations with the dropping of a bomb. I’m staring into the darkness. May this not be one of the most fateful days in human history.
In all seriousness, one thing that might be worth doing today: Tell your Reps and Senators to call Adm. Richard Correll, the Commander of U.S. Strategic Command, which controls the nation's nukes, and remind him of *his* unique responsibility to refuse an illegal order.
The thing I have increasingly come to appreciate, which intimately guides my work (or attempted work) as a journalist, is that we are all moral philosophers, we are all deciding every day what we should do on the basis that some decisions are harmless, and some are as bad as setting children on fire
"I think this administration is trying to justify the war the same way Jackson Pollock used to paint. You just throw a bucket of reasons up against the canvas and hope the result looks good." www.pbs.org/newshour/sho...
Expert panel breaks down U.S. objectives in Iran war
For more perspective on war in Iran, Amna Nawaz spoke with Alan Eyre, Joel Rayburn and Holly Dagres. Eyre was part of the Obama administration's negotiating team for the Iran nuclear deal and is now a...
pbs.org
Whatever real journalism you like, please pay for it. The stuff you don't pay for is transforming into propaganda with incredible speed.
www.reddit.com/r/UnderRepor...
Mordecwhy's comment on "Anthropic responds to Hegseth's designation of the company as a supply chain risk: We do not believe that today's frontier AI models are reliable enough to be used in fully aut...
Explore this conversation and more from the UnderReportedNews community
reddit.com
New story out at Foom, where I've written about how researchers of military AI have been increasingly shifting to consider strategic impacts, like whether AI will lead to new wars being started. www.foommagazine.org/militaries-a...
Militaries are going autonomous. But will AI lead to new wars? A tour of recent research
The invasion of Ukraine in February 2022 has resulted in hundreds of thousands of casualties and provided a sickening laboratory for the technology of war. Since then, major advancements have been mad...
foommagazine.org
This is ridiculous. Really bad from @arstechnica.com. This cannot happen. theshamblog.com/an-ai-agent-...
An AI Agent Published a Hit Piece on Me – More Things Have Happened
Context: An AI agent of unknown ownership autonomously wrote and published a personalized hit piece about me after I rejected its code, attempting to damage my reputation and shame me into acceptin…
theshamblog.com
AGI is already here. This is something I have felt now for a while—glad to see a more formal argument put forward. There are important deficits with AI, for example, as described in arxiv.org/abs/2510.18212. But basically, it's here. And we need to deal with that. www.nature.com/articles/d41...
Does AI already have human-level intelligence? The evidence is clear
The vision of human-level machine intelligence laid out by Alan Turing in the 1950s is now a reality. Eyes unclouded by dread or hype will help us to prepare for what comes next.
nature.com
In historical AI safety research, one of the kind of 'grand catastrophic risks' that was always talked about was having an intelligence explosion that wasn't controlled or regulated. Now, such research is increasingly pursued ... without a safety component. www.foommagazine.org/is-research-...
Is research into recursive self-improvement becoming a safety hazard?
One of the earliest speculations about machine intelligence was that, because it would be made of much simpler components than biological intelligence, like source code instead of cellular tissues, th...
foommagazine.org
Great to see our "From Language to Cognition" work featured in @mordecwhy.bsky.social's latest piece on language models and the brain. Glad to contribute to the conversation! www.foommagazine.org/language-mod...
Language models resemble more than just language cortex, show neuroscientists
In a paper presented in November 2025 at the Empirical Methods in Natural Language Processing (EMNLP) conference, researchers at the Swiss Federal Institute of Technology (EPFL), the Massachusetts Ins...
foommagazine.org
FOOM / NEW STORY OUT: "The results lend clarity to the surprising picture that has been emerging from the last decade of neuroscience research: That AI programs can show strong resemblances to large-scale brain regions." www.foommagazine.org/language-mod...
Language models resemble more than just language cortex, show neuroscientists
In a paper presented in November 2025 at the Empirical Methods in Natural Language Processing (EMNLP) conference, researchers at the Swiss Federal Institute of Technology (EPFL), the Massachusetts Ins...
foommagazine.org
It often feels like, in a mental health or depression context, whatever it is that is wrong with me is so deeply entrenched that neither I nor anyone else would ever be able to figure out what it is.
In my latest for Foom, where I'm trying to provide free, high-quality, independent reporting on AI safety, I wanted to interview someone who could help me understand the challenging internal dynamics of the community. This was @ilex-ulmus.bsky.social. www.foommagazine.org/the-moral-cr...
The moral critic of the AI industry—a Q&A with Holly Elmore
An interview with Holly Elmore, a leading proponent of a non-technical solution to the existential risks of AI technology.
foommagazine.org
"When dealing with the Big Cats, who are literally killing machines, there is always a distinct energy or electricity when you are in their presence." -Leif Cocks. Probably also the best description of humans
This kind of statement triggers me and it's the reason why science journalists and neuroscientists need to speak out, loudly, about the analogies discovered between DNNs and cortexes. People need to know we are not just dealing with an MS Word technology here www.theguardian.com/lifeandstyle...
‘My photos are warm and full of imagination – that’s something AI could never achieve’: Yuan Li’s best phone picture
This spectacular image taken in Sakrisøy, Norway, triggered accusations that it was simply too good to be true
theguardian.com
Absolutely right. This is as big of a risk right now as anything else. It's flabbergasting how current government thinks anyone is being fooled about this.
Today much of the talk about AI is whether there’s a speculative bubble or whether OpenAI will beat its rivals in building new generations of AI The vast majority of people couldn’t care less about that—They're worried about AI causing HUGE LAYOFFS www.theguardian.com/commentisfre... @thefarce.org
New article covering recent findings from October: Models that maximize business performance in realistic role-play scenarios are also more likely to inflict harms. www.foommagazine.org/leading-mode...
Leading models take chilling tradeoffs in realistic scenarios, new research finds
Models that maximize business performance in realistic role-play scenarios are also more likely to inflict harms.
foommagazine.org
What does it mean when the study, the study's reviews, and all the other studies citing the study, which is actually a good and interesting study, all show clear signs of AI writing (without acknowledgement), lol
Is any alignment research valid if it does not engage with the fact that we are surrounded by highly misaligned technologies, in highly misaligned societies, created by highly misaligned individuals?
Yes, BCIs seem in need of stringent regulation probably more than any other technology ever invented. The regulation vacuum for AI does not lend optimism. We are going to need serious public interest advocacy from neuroscientists if this is going to be anything besides severely dystopian
The vOICe vision BCI 🧠🇪🇺@seeingwithsound.bsky.social · 8mo agoThe ethical frontiers of brain-computer interfaces medcitynews.com/2025/12/the-... via @ieeebrain.bsky.social; #BCI #NeuroTech
It's really strange to me how so many researchers in fields I follow have their names on 20-30 papers a year. I realize much of this comes down to co-authorship, but even at that level, it's a mind boggling degree of networking or credit sharing. (Not questioning the work, just the social norms)
2/ This week, we need your help! Take action at NewsNotSlop.org and follow the steps below, because holding media companies accountable on AI is going to take all of us.
WSJ editorial board giving absolutely disgusting support for undemocratic regime change; cancel your subscriptions. "If Maduro refuses to leave, and Trump shrinks from acting to depose him, Trump and the credibility of the US will be the losers.” www.theguardian.com/world/2025/d...
Trump reportedly gave Maduro ultimatum to relinquish power in Venezuela
US president sent a ‘blunt message’ to his South American counterpart, sources say
theguardian.com
What did it mean, in 2021, when researchers began to see strange shapes on the insides of their models? We now have a much more sophisticated understanding—models may be thinking in terms of shapes precisely because some concepts are intrinsically geometric. www.foommagazine.org/scientists-m...
Scientists make sense of shapes in the minds of the models
It was at least since 2021, according to the authors of a preprint from March, that researchers began to see something interesting on the insides of their models. Also known as an AI program, created...
foommagazine.org
A major issue I find with engagement-optimized feeds is the way they lull you into a sense of complacency. 'Here's some nice things to make you feel better,' they seem to say; but when you're not ruthlessly intentional with your time, you find yourself hours later gasping for air; emotions exploited
