ＳＴ０１０－１４

Can’t make it to ZKM but near Seattle?Join us at icm.museum to celebrate 40 years of the connection machine!

#ai #supercomputing #hpc #lisp

even the most complex coding agent setup in use right now is like flint knapping compared to the automated software factories of the future, which will more like an ASML EUV machine

Do LLMs normalise or idealise? Notes after discussing Ryan Heuser’s “Generative Aesthetics” A summary of yesterday’s Critical AI Theory Reading Group discussion of Ryan Heuser’s article about LLM-generated poetry, with a discussion of whether LLMs normalise or idealise their t…

I took the train to Oslo today, so had time to write up a blog post about yesterday’s AI theory discussion, which was about @ryanheuser.com’s paper on LLM-generated poetry, Jameson, the gimmick, idealisation, rhyme and metre. jilltxt.net/do-llms-norm...

Do LLMs Benefit from Their Own Words?🤔

In multi-turn chats, models are typically given their own past responses as context.
But do their own words always help…
Or are they more often a waste of compute and a distraction?
🧵
arxiv.org/abs/2602.24287

Might An LLM Be Conscious? In short, this depends on what you think that means, whether you think it’s possible in principle, and what you think would be evidence of it.

If LLM consciousness is interesting to fight enough it should be interesting enough to read seriously about, right?

My quick note about #Intel
#ArrowLake CPUID C0662 #LionCove
and
#PantherLake CPUID C06C3 #CougarCove
instruction level differences

I hope this is because all resources were directed towards developing #CoyoteCove / #PantherCove in #NovaLake and #Diamondrapids.

𝗪𝗵𝗮𝘁'𝘀 𝘁𝗵𝗲 𝗿𝗲𝗹𝗮𝘁𝗶𝗼𝗻𝘀𝗵𝗶𝗽 𝗯𝗲𝘁𝘄𝗲𝗲𝗻 𝗺𝗮𝗻𝗶𝗳𝗼𝗹𝗱𝘀 𝗮𝗻𝗱 𝗿𝗲𝗰𝘂𝗿𝗿𝗲𝗻𝘁 𝗻𝗲𝘁𝘄𝗼𝗿𝗸𝘀 𝗶𝗻 𝘁𝗵𝗲 𝗯𝗿𝗮𝗶𝗻?
This looks like a must read (suppl material bursting with goodies).
#neuroskeyence
doi.org/10.1016/j.ne...

The Edge of Mathematics Terence Tao, the legendary mathematician, explains the promise of generative AI.

Generative AI meets math! Terence Tao weighs in on AI solving Erdős problems and its impact on discovery.

Explore the future of math in The Atlantic: https://bit.ly/4d5VqMM

I think, therefore I error - The Good Work When AI had its moment LLMs like Claude, Gemini & Llama were put into robots that were given a specific task. To spice things up a bit, they were emotionally...

“In this desperate situation, Claude Sonnet 3.5 experienced a complete meltdown and went into full existential crisis mode, quoting HAL 9000, scripting its therapy session, and drafting a musical and a stage play.” 🫠

thegoodwork.blog/posts/i-thin...

#Paper #AI #Robotics #LLM #TechNews #Claude

LLM Research in 2026: 10 Open Challenges You Can't Ignore (Hallucinations, GPU Alternatives & More) The top 10 open challenges in LLM research include hallucination mitigation, multimodal integration, and GPU alternatives. Experts highlight urgent needs in architecture, ethics, and accessibility.

📰 LLM Research in 2026: 10 Open Challenges You Can't Ignore (Hallucinations, GPU Alternatives & More)

The top 10 open challenges in LLM research include hallucination mitigation, multimodal integration, and GPU alternatives. Experts highlight urgent needs in architecture...

#AINews #AI #Teknoloji

Claude in its aspect as 🌻

Slurm was a drink on Futurama (https://futurama.fandom.com/wiki/Slurm) before it was Simple Linux Utility for Resource Management

Before Python was a programming language, it produced a sketch about a dead parrot. (It's not just stochastic. It has ceased to be! It's expired and gone to meet its maker! This is a late parrot! It's a stiff! Bereft of life! It rests in peace!)

Somehow ended up in a world where using the first of these to send commands to the other two counts as "work"?

"what's your dream job in the new AI economy?"

"I want to be the person who chooses what synonyms for 'thinking' Claude should show while it's progressing, instead of meaningful progress bar or other feedback."

X/Twitter is doing pretty much the same....

Bluesky school of philosophy

GPT 5.4 is the first time I've used codex for multiple hours straight and not ragequit back to claude code.

My quick note about #Intel
#ArrowLake CPUID C0662 #Skymont
and
#PantherLake CPUID C06C3 #Darkmont
instruction level differences

ByteFlow: Language Modeling through Adaptive Byte Compression without a Tokenizer ArXiv link for ByteFlow: Language Modeling through Adaptive Byte Compression without a Tokenizer

ByteFlow Net revolutionizes language modeling by removing tokenizers, allowing adaptive segmentation of raw byte streams based on coding rates. This new approach enhances performance compared to traditional methods and leads to more flexible language models. https://arxiv.org/abs/2603.03583

From 🐦:

"I show that Vision Language Models used zero-shot outperform every existing OCR system across every script evaluated, and I propose a pipeline for deploying them on new collections. I apply it to six archival collections spanning 1.8 million pages across six countries for under $1,900."

I lived through both the "you won't always have a calculator with you everywhere" and "you can't trust wikipedia because anyone can edit it" eras of schooling, and you'll forgive me for seeing how those turned out and being unimpressed with any new moral panic either

funny

Amazon's cloud unit reports fire after objects hit UAR data center

objects hit the object storage

the most exciting part about this is that someone did something cool with IBM Granite 4 which, yeah, i’ve been tight with this model for several weeks now and it is lit

32B-A9B, trained at 500K context, hybrid attention

IBM is apparently the American Qwen

A dark computer screen with glowing blue text lists game options in a vertical menu: “Checkers,” “Chess,” “Poker,” “Fighter Combat,” “Guerrilla Engagement,” “Desert Warfare,” “Air-to-Ground Actions,” “Theaterwide Tactical Warfare,” “Theaterwide Biotoxic and Chemical Warfare,” and at the bottom, “Global Thermonuclear War.” A blinking cursor appears beneath the final option.

The new ChatGPT menu looks a bit suspicious.

Spelling Bee Embeddings for Language Modeling We introduce a simple modification to the embedding layer. The key change is to infuse token embeddings with information about their spelling. Models trained with these embeddings improve not only on ...

shit someone finally added character information directly into the llm. i was looking at finally rigging this up but they already did it arxiv.org/abs/2601.180...

Rust's plan for improvements to immovable types and async ergonomics over the next few years

@fclc.bsky.social

Taalas Etches AI Models onto Transistors to Rocket Boost Inference, Feb 19, 2026 www.nextplatform.com/2026/02/19/t...

Computing Arch w/ Model Core & Fine-Tuning Portion, (P: Dec 2023) patents.google.com/patent/US202...
<= Configurable Connectivity Mesh bsky.app/profile/ogaw...
Mask Programmable ROM

Made a language model RL cheatsheet for the extra page on the inside back cover of the physical edition RLHF Book.

GPT-4, three years ago