'algorithm' has 2 meanings: 1. a step-by-step procedure with finite description that halts in finite time 2. an algorithm *for* a function/problem LLMs can fail for both definitions. The first because of human-in-the-loop, the second because they are heuristics at best (no guarantees)

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

Yes absolutely. Worst so far is opening a PDF PhD thesis to examine “THIS LOOKS LIKE A LONG DOCUMENT. WOULD YOU LIKE A SUMMARY?” Seriously, 4 years of the student’s life & research…

and the irony of this popping up as i try to read papers on the reproducibility crisis in machine learning research, exacerbated by llms

i hate it @wileyonlinelibrary.bsky.social, please kill this

i love how every truck and train company is called Trans Something

Humans deliberately using AI to harm people is a concrete danger, and that is what deserves serious attention. Not absurd "ai takes over the world" scenarios.

After a few rounds of editing, there's really nothing left to correct in a one-page document. But Gemini starts clutching its pearls about how I should include my middle initial in the sign-off, and my failure to do this is *a breach of professional norms.* (This is hardly the worst I've seen.)

Defy Category itself.

Playing Half-Life: Alyx is still thrilling on the Steam Frame, but it does make you confront the idea that a VR game this ambitious in scope will never come out again. aftermath.site/half-life-alyx...

Will There Ever Be Another Half-Life: Alyx?

Half-Life: Alyx remains the big budget killer app of VR. But playing it again makes you wonder if anyone will ever be allowed to match that ambition.

aftermath.site

The latest “pain axis” research, to me, seems similar to Anthropic’s “functional emotion” research in that it basically falls out of base models being good at writing text that’s coherent at long context- *some* representation of the mental state of the writer must exist in order to do what it does

they can now solve persona 6

I'm basically the world expert on AI

dQwen3.5: Hybrid-Attention Diffusion Language Models arxiv.org/abs/2609.20751 dQwen3.5-2B huggingface.co/UT-IFML/dQwe... dQwen3.5-4B huggingface.co/UT-IFML/dQwe... dQwen3.5-9B huggingface.co/UT-IFML/dQwe...

dQwen3.5: Hybrid-Attention Diffusion Language Models

Adapting a pretrained autoregressive (AR) model is a cost-efficient route to a diffusion language model (DLM). While nearly all such adaptations start from a full-attention transformer, AR modeling ha...

arxiv.org

good morning

dude just fucking hates moderators

Slavoj Žižek has my dream job. I watched him speak in front of a crowd for 45 minutes uninterrupted about whatever he felt like while the host watched helplessly. when he finally finished she said "wise words" and he said "I HATE WISDOM" then started talking again

There is a through line between a few companies who think they’re building superintelligences, who oddly don’t seem to care much about their consumer or business facing products, trying to form a cartel, Anthropic building a wet lab, and some gray tribe beliefs about the future of democracy.

Neural networks did not work UNTIL they got all that data. Raw data is useless without immense human labour to clean and store it in massive data centres. GPUs which are also required need rare earth minerals. On and on. AI only works because we're dedicating all of a massive industry to it.

Olivia Guest · Ολίβια Γκεστ@olivia.science · 2w ago

You haven't bothered to watch my video, which is fine, but you'd learn none of that is recent (see slide). This idea of pretending it's new when it's always been like that is part of the problem. You have to really be more careful as you keep falling into their traps. bsky.app/profile/oliv...

We can destroy Tilly Norwood by having One journalist ask it’s opinion on Palestine and Israel

honestly some of the most apocalyptic impacts of "AI" will be shit like this, and most of the doomsday dipshits predicting SkyNet can't be bothered to notice or care

UnitedHealth uses AI model with 90% error rate to deny care, lawsuit alleges

For the largest health insurer in the US, AI's error rate is like a feature, not a bug.

arstechnica.com

Nicholas Grossman@nicholasgrossman.bsky.social · 2d ago

There have always been humans who thought: 1) The world will end soon 2) It will happen in a cool, exciting way 3) It’ll happen to me personally 4) I’ll be a martyred prophet, or a warlord, or otherwise special 5) People who don’t see this are lesser “Sexy apocalypse” is a great phrase for that

If the value of protecting endangered species isn't obvious to you, imagine instead that it is protecting the last ruin of an ancient civilization from being bulldozed for something not particularly important. Imagine the ruin is 30 million years old.

Ryan Moulton@moultano.bsky.social · 3d ago

This is insane.

A serious weakness in the “sufficiently intelligent AI inevitably wages war on humanity” story is that it needs to explain why conquest beats purchasing, contracting, and participating.

There is no AI kill switch that is not an Internet Kill Switch. They're discussing having in the US, what they have in China and Iran.

Techmeme@techmeme.com · 24h ago

Experts say AI kill-switch legislation is far harder to implement than lawmakers assume, warning a rogue AI could actively try to dismantle the mechanism itself (New York Times) Main Link | Techmeme Permalink

www.cnn.com/2026/09/19/b...

Lawsuit says Anthropic, OpenAI, SpaceXAI and Google made illegal agreement on AI slowdown | CNN Business

A new lawsuit claims Anthropic, OpenAI, SpaceXAI and Google made an illegal deal to slow the pace of their respective AI development.

cnn.com

Notably, this is good old fashioned human greed, and not fantasy literature reasoned from thinking very hard with a very big human brain.

Notably, giving the cultists a cartel is the opposite of a solution