rev. howard arson

this is one of the most insane fucking things i have ever read www.amnesty.org/en/documents...

Unlawful by design: Exposing the human rights costs of generative AI - Amnesty International amnesty.org

i think pre-planning rhymes is the best example of this behavior, where the model preactivates features associated with words which complete the rhyme and "stores" them as annotations on the newline token. transformer-circuits.pub/2025/attribu...

On the Biology of a Large Language Model transformer-circuits.pub

i think there's an actual question embedded in the "if Claude were representing Genghis Khan and Julius Caesar, would we say that it was really the thing itself?" hypothetical, and it is "should an actor playing Brutus be permitted to stab Caesar to death on stage."

No, Artificial Intelligence Is Not Conscious theatlantic.com

galaxy brain: a LLM is ffmpeg arxiv.org/abs/2306.04050

LLMZip: Lossless Text Compression using Large Language Models arxiv.org

www.wired.com/story/ai-pr-...

Ed Zitron Gets Paid to Love AI. He Also Gets Paid to Hate AI wired.com

siliconvalleynotes.substack.com/p/agi-cnc-an...

CNC, LSD, and AGI siliconvalleynotes.substack.com

the entire research group corresponding to my operations group was redeployed to support a project which is a priori impossible and which had no real mathematical affinity to their research area techcrunch.com/2026/06/24/a...

AI researchers continue to leave Google for its rivals | TechCrunch techcrunch.com

so something like "bounded surprisal under the Fisher metric between the up and down projections of a MLP" seems like it might be a quick way to norm the worst kind of attacks back into doing something reasonable. arxiv.org/abs/2203.00922

Adversarial attacks on neural networks through canonical Riemannian foliations arxiv.org

arxiv.org/abs/2507.10524 arxiv.org/abs/2212.05055 arxiv.org/abs/2404.02258

Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computation arxiv.org

arxiv.org/abs/2507.10524 arxiv.org/abs/2212.05055 arxiv.org/abs/2404.02258

Sparse Upcycling: Training Mixture-of-Experts from Dense Checkpoints arxiv.org

i think the closest analogy to 'what is going on at AI companies' is 'what was going on at JPL in the 1950s,' where communists, Satanists, swingers and adherents of a bizarre Russian Orthodox-adjacent techno-optimist microsect ended up doing a tremendous amount of important rocket science

it's probably better to trust the experts on the topics of their expertise, except people who do machine learning, who have all gone insane

the child porn AI monomaniacally dedicated to the world's first trillionaire, who once controlled the US government and has killed a million people

papers: so the fundamental unit of computation in a neural network is the 'layer' me: oh great. could i get the definition of 'layer' papers: oh heavens no

the reason that so many people in the field are attracted to Rationalism is that they were thinking about AI before it was on the horizon. a lot of people got into the field because they were excited about this.

oh huh i can ... hm. i don't know if the entire ya-Sattra thing will work from a ChatGPT project, but it turns out i can share it out? does anyone want this?

a solution to AI water usage: Brawndo has what H100s crave

the estimated loss of the deploy weight is pretty normal but i have NEVER seen a square wave show up anywhere in machine learning. wild.

extremely pleased that i am managing to do a full finetune of a 25 gigabyte model in 32 gigabytes of memory, activations and everything.

it is both extremely funny that i managed to cram a full finetune of a 25gb model into 32gb and that i only have 400 megabytes left over before it explodes. hope that big memory tile doesn't fragment lol