Join free with your email
Free. No spam, ever — we'll never share your email address and you can opt out at any time. Já tem conta? Entre
Directory member with public evidence across AI research.
No, OpenAI's new magic models did not autonomously hack Huggingface. Per OpenAI's PR blog post (https://openai.com/index/hugging-face-model-evaluation-security-incident/): "This incident occurred during an internal evaluation which prompts models to pursue advanced exploitatio…
I know it’s not a popular position but I really don’t see mass unemployment from generative AI soon because of structural and technical features of the tech. Gift link where I explain the reasons. It’s distraction when AI will really cause many other big problems. www.nytimes.…
“Norway imposes near ban on AI in elementary school | Reuters” https://www.reuters.com/technology/norway-imposes-near-ban-ai-elementary-school-2026-06-19/ >
“In the report, the auditor general points out multiple concerning examples of mistakes in those summaries that could have a direct and negative impact on a patient’s subsequent care.”
“AI Isn't Management. Try Explaining That to Matthew Prince” https://www.programmablemutter.com/p/ai-isnt-management-try-explaining > At last we have created the Corporation That Eliminates Middle Managers from classic management text Don't Eliminate the Middle Managers!
“How to use AI doom marketing to dupe the media and rake in billions in 10 easy steps” https://www.bloodinthemachine.com/p/how-to-use-ai-doom-marketing-to-dupe > The Mythos campaign is nothing less than the ne plus ultra of AI ‘doom marketing’
> Without significant changes to how these systems are developed and deployed, they will continue to undermine international human rights laws and standards. www. amnesty.org/en/documents/pol40/0996/2026/en/ This is a pretty good report overall. Hard to disagree with any of th…
“How to talk about "AI" without adding to the anthropomorphization • Buttondown” https://buttondown.com/maiht3k/archive/how-to-talk-about-ai-without-adding-to-the/
“How Anthropic used AI ethics slop to play the pope and eclipse OpenAI” https://www.bloodinthemachine.com/p/how-anthropic-used-its-ai-ethicslop > Anthropic’s opportunism leaves a stain on the entire encyclical
“"Useful" is not sufficient” https://tante.cc/2026/07/15/useful-is-not-sufficient/ > Usefulness is not an objective, fully abstract property: A thing is not useful for everyone or in general but mostly just for specific people in specific contexts. And useful never comes for f…
“Potentially Useful | Hating AI in 2026” https://www.eamoncaddigan.net/posts/ai-in-2026/ > We can point out all of the harms being done by the technology right now, and we can speak to the additional harms that will be done if things continue like this, but our concerns are sp…
One of my worst speech technology nightmares has been happening right in my own city! In the latest Mystery AI Hype Theater 3000, @alexhanna.bsky.social and I dig into it: www.buzzsprout.com/2126417/epis...
People keep using the "can't put a genie back in the bottle" metaphor for "AI", even though in the stories they always get back in and current "AI" is more of a "we're spending trillions of dollars and the energy output of a continent trying to keep this bastard out of the bottle" situation
Current “AI” from my perspective is less like a technological breakthrough—“the genie is out of the bottle”—and more a research fusion reactor: no matter how much energy and money you throw at the thing, nothing changes the fact that it costs more energy than it produces
Just once I’d like to see journalists and politicians show at least a grain of scepticism about the claims being made by “AI” vendors. Just once. Just because anthropic styles their reports to look like scientific or academic papers doesn’t make them sound or credible.
Only just now realising that the wish fulfilment of having a dedicated slave or polite servant might play a bigger role in the popularity of chatbots than I expected
The "AI Compass" quiz, though broadly accurate in terms of my own results ("Luddite (affectionate)") is inherently a piece of pro-"AI" propaganda because it is one long exercise in obfuscating the differences between Machine Learning, LLMs, diffusion models, and other forms of statistical computing
One of the annoying consequences of LLMs is that many people now seem eager to dismiss any and all writing that people enjoy as “obviously LLM-generated”, just because it uses em-dashes or writing structures with decades if not centuries of history in English.
Finding productive uses for LLMs while at the same time doing nothing to mitigate the numerous harms makes things WORSE not better. Responding to “AI” critique by saying that they have their uses and with work we can make them genuinely productive is not a rational response. It’s accelerationism
This is your regular reminder that even if you “dispassionately” assess LLM tools purely in economic terms, they are a complete and utter failure. Massive costs and capital expenses (chips, training, datacentres) that deprecate rapidly, all for a negligible boost in economic productivity
The problem with positioning yourself as a sensible and reasonable moderate user of LLM software development tools is it means you’ll be the first up against the metaphorical wall if either the anti- or pro- sides truly gain the upper hand.
It doesn’t really matter that much how the boosters end up defining “slop” or “vibecoding” or whatever new term they come up with because they’ll all ultimately become synonyms with “AI” and the ship has sailed on that meaning anything other than “garbage” in the public’s mind.
Center = Baldur Bjarnason. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.
