ok, as promised my take (part 1 of N) on (scholarly) reading skills:
> so you must understand that you are entering another world. Think of it a little like stepping into another person’s house. You cannot open random doors and expect the kitchen to be where yours is.
artcraft looks really impressive, like the math stuff looked really impressive before it turned out that openai stole results from a mathematician or the vulndev stuff sounded really impressive before it turned out that mythos was 95% false positives or the coding efficiency gains looked really impressive before amazon had a day-long outage and uber wiped out its budget in 3 months. so please do not talk to me about it just yet.
2. But Anthropic researchers say they see patterns that suggest that Claude can.
The employees at Anthropic are telling themselves stories based on the words they input into the software and the words that come back out.
It's the job of the media to scrutinize such statements and decide whether there's a story here at all (maybe the story is Anthropic researchers have fallen into a fantasy-land and brought many policymakers with them), not to just platform them.
>>
1. Can a language model feel pain?
Clearly not. Large language models are computer software and they can no more suffer than any other human-created artifact can.
>>
#EU drops sponsorship of #disinformation conference at #Trump officials’ request
🖊️ Aisha Down in London, Joseph Gedeon in Washington and Jennifer Rankin in Brussels
#Lithuania and #Canada were also pressured to withdraw, and days before #Disinfo2026, logos were wiped off website
https://www.theguardian.com/us-news/2026/oct/07/trump-eu-disinformation-conference
The people I agree most with on the likely technical trajectory of "AI" tend to trend much more blackpilled than me on societal implications, and I don't really know what to do about that.
Like, it's bad, yeah, but anything that can't go on will stop. And if every decent software developer (used in the broadest possible sense), to take recent example, preemptively leaves the industry, who will be there to right the ship or rebuild, when the {hype,financial} bubble pops?
Conversely, am I saying the developers should break themselves upon the rocks that are the current, common "AI" adoption stance in most companies? No, I'm saying decent developers should find ways to start building resilient, resistant organizations now.
I'm in one. We are not pushing "AI" adoption. We are doing fine. We have many customees apprecisting the things we do, which no LLM could do. The work still exists, let's build a foundation for the future there.
This too shall pass, and then what…?
We can agree to disagree. You can think that you've got enough subjective controls on your subjective assessment of your own productivity that you're making the best judgement available in your situation. (Maybe you should think a bit more about the ethics, but, okay, that's a separate conversation, and you might also have an answer for that.)
But please stop telling me that this hypothesis is "obviously stupid" without sharing _any_ evidence that is even the slightest bit incongruent with it.
Really tired of having this conversation over and over again.
A: "LLMs provide the illusion of productivity, by making you *feel* productive while wasting enough time that they're net negative."
B: "Come on. That's not a serious position any more. Maybe 3 years ago. Not today. Obviously they work."
A: "How do you know?"
B: "It's obvious."
A: "… because *you* use them, and they make *you* feel productive?"
B: "Are you calling me DELUSIONAL? Everybody who disagrees with you is just CRAZY?"
Everyone needs to read this. Reality is worse than anything we can say about these people in a few words, media appearances, op-eds, or other conversations. Read the whole thing.
🔇 🔇 🔇 Reminder: These are the people advising Bernie Sanders on "AI" policy.
Link: https://skywriter.blue/@segyges.bsky.social/3mvom4b4dn22q
We call failures of this binding Loopjacking: a human approves what they understand as operation A, while the implementation uses that decision for a materially different operation B.
Friendos we call that lying and it has been around for literally forever.
Researchers analyzed the epistemic diversity of LLMs (meaning: how many different positions/kinds of information do they produce for a given input). Result:
"every system is less diverse than a search baseline. This gap is not uniform: RAG can improve diversity, while large models are counterintuitively less diverse than smaller ones. Moreover, LLM parametric knowledge systematically reflects English over local-language knowledge for country specific topics."
Tech bros are like “I built this!”
Dude, you didn’t build diddly.
The labour your capital captured built it; you just took credit and made sure the resulting work was kept out of the commons. The only thing you did was enclose human potential to benefit the few instead of the many.
This is what capitalists do: they reduce human potential into a resource to be farmed within a zero-sum game, thus stealing from all the possible other futures of humanity.
A lot of her reporting in the last couple of years has been declining in quality and less and less interested in the truth, but her recent output has been full-on defensiveness about AI, so, it's all of a piece. When I've been critical in the past I've tried to focus on the work itself and not ad-hominems but when people keep telling you who they are eventually you've got to believe them.
Taylor Lorenz is preparing for her "why I left the left" right-wing grifter pivot. Do not trust her. https://www.thefp.com/p/two-drinks-with-taylor-lorenz-savodnik
(If you do not know "the free press", it is Bari Weiss's right-wing false-balance propaganda outlet, so talking to them *at all* is already suspicious. Talking to them about how you "wouldn’t say [you're] particularly woke" is a dog whistle so loud my ears are bleeding.)
New from @timnitGebru and me:
"It is high time for us to regulate usage of systems sold as “AI”, not because they are magical machines, but because they are error-prone products that shouldn’t be used in high stakes scenarios."
RE: https://infosec.exchange/@david_chisnall/117358990093685067
100%. The whole "terrorists can use an LLM to build a bomb" is a dumb movie plot threat. If you want to hurt people, lack of knowledge is not your problem. And nobody needs to develop a new chemical to do terrorism. All you need is readily available.
This is just "AI" labs creating imaginary issues to distract us from the real harm they cause.
Even if the code quality hadn’t declined overall, LLM users would still make the worst collaborators.
“Imagining the Car but not the Traffic Jam”
https://azhdarchid.com/imagining-the-car-but-not-the-traffic-jam/
> In most fields where the consequences for error are serious, human performance is not 99% accuracy or 99.9% accuracy; it's 99.99% accuracy or better.
(Via @olivia )
okay so since people keep saying that AI training is fair use and that you need to be able to spot plagiarism based on the contents of a repository: no. please read https://ansuz.sooke.bc.ca/entry/23