[LangExtract](https://developers.googleblog.com/en/introducing-langextract-a-gemini-powered-information-extraction-library/) has got me curious, but I don't get what makes it different from a [spacy-llm/prodigy](https://prodi.gy/docs/large-language-models) setup. Is it just that I am spared the effort of chunking long input and/or constructing output JSON from entities and offsets by writing the corresponding python code myself?...

Ah, one more difference is that langextract is #OpenSource whereas prodigy is not (?). (On the other hand, prodigy has a better integration with a correction+training workflow.)

> #GAIA is a #generativeAI application designed to run local, private #LLMs on #Windows PCs and is optimized for #AMDRyzen #AI hardware (#AMD #RyzenAI300 Series Processors). This integration allows for faster, more efficient processing – i.e. lower power– while keeping your data local and secure. On #Ryzen AI PCs, GAIA interacts with the #NPU and iGPU to run models seamlessly by using the open-source #Lemonade (#LLMAid) SDK from ONNX TurnkeyML for #LLM inference.

https://www.amd.com/en/developer/resources/technical-articles/gaia-an-open-source-project-from-amd-for-running-local-llms-on-ryzen-ai.html

*What's the "best way to do this"? Maybe not to do it at all? But people will, because it's easy. #LLM #designfictions

https://medium.com/@bruces/the-story-of-filboid-studge-or-rebuilding-imaginary-futures-4f7e954f84f5

“These findings suggest a scenario where machines, originally trained on human data and subsequently exhibiting their own cultural traits, can, in turn, measurably reshape human culture.”
Do we really want AI influencing our vocabularies? What else will it affect?
#AI #LLM

https://arxiv.org/html/2409.01754v3

This MIT study suggests to me that the race for AI is one against time. If it's accurate (This was a small sample with a larger study in the works.) the people using Ai (presumably including those working to build AI for Microsoft, Meta, etc.) are undergoing rapidly deteriorating brain function.
This also puts into question plans to use AI in schools.
https://www.media.mit.edu/publications/your-brain-on-chatgpt/
#MIT #AI #Programmers #LLM #Cognitive #Education

Tell you what, #LLM operators. I'll make you a deal.

You're free to ignore our robots.txt files so long as you give us the right to pour a few hundred gallons of Diet Coke into your server racks.

https://blog.cloudflare.com/perplexity-is-using-stealth-undeclared-crawlers-to-evade-website-no-crawl-directives/

Newelle is a desktop AI assistant for Linux, providing a native GTK front-end to cloud and local LLMs. It features voice chat, long-term memory and extensions.

https://www.omgubuntu.co.uk/2025/08/newelle-ai-assistant-ubuntu-linux-desktop?v1

I feel like I'm in this perfect sweet-spot for #vibecoding; know enough about #programming to spot mistakes, and enough about #software design to communicate requirements simply, but also too lazy to ever be fucked to do it myself. I'm suddenly finding all these opportunities to automate stupid shit I've been doing by hand because I was too lazy to write a script for it. It's really leveled the Adderall playing field.

Man, #ClaudeSonnet4 is so good at #programming, I wish it wasn't so expensive.

AI haters: "#AI destroys your critical thinking skills"

Also AI haters: "you have to be critical of everything that #AI says because it makes things up"

So it destroys your ability to do the thing it makes you do?

> We see continued evidence that #Perplexity is repeatedly modifying their user agent and changing their source ASNs to hide their crawling activity, as well as ignoring…robots.txt files.

https://blog.cloudflare.com/perplexity-is-using-stealth-undeclared-crawlers-to-evade-website-no-crawl-directives/

¡Buena noticia!

https://proton.me/blog/lumo-ai
#DigitalSovereignty #Proton #AI #LLM #europe

"LLMs Are Not The Path To AGI:

Apple’s conclusion is the same conclusion I came to very early on in my experimentation with LLMs. … the position that LLMs are not, & never will be AGI. My position has remained completely unchanged.

… Of course, language translation is not AI, Siri is not AI, image identification & classification is not AI, the only acceptable form of AI is shoving some half-baked LLM chatbot into somewhere it doesn’t belong."

#AI #LLM
From @malwaretech https://infosec.exchange/@malwaretech/114969567431656774

Proclaiming that #AI is a bubble / is useless / is not real etc. because you had recent poor experiences with an #LLM/chatbot is a bit like saying electricity will never be useful because your electric toaster burnt your bread this morning.

LameHug Malware uses AI LLM to craft Windows Data-Theft Commands in Real-Time.

LameHug was discovered by Ukraine’s national cyber incident response team [CERT-UA] and attributed the attacks to Russian state-backed threat group APT28 [a.k.a. Sednit, Sofacy, Pawn Storm, Fancy Bear, STRONTIUM, Tsar Team, Forest Blizzard].

https://cert.gov.ua/article/6284730

This article raises an issue about the AI boom that I haven't seen much comment on.

More AI means more data centres.

In some cases, they'll go on land that could be used for housing.

For example, this one will go on land approved for 1400 new apartments: https://www.realcommercial.com.au/news/macquarie-technology-forks-out-240m-to-pick-up-data-centre-site-from-holdmark

"Macquarie Technology Group has added fuel to the country’s AI-driven data centre boom by snapping up a sprawling site in Sydney’s north once earmarked for high-rise apartments.

"ASX-listed Macquarie Technology will pay $240m for the Macquarie Park site that has been caught up in the complicated state planning system for almost a decade. The Talavera Road property, which it will buy via an option agreement, is near another data centre development it has in the suburb.

"The new property had been long earmarked for about 1400 apartments with Macquarie Technology saying it was bought from an established developer.

"Its removal from the pipeline will put pressure on the area’s capacity to contribute to housing targets, with the Albanese government sticking to plans for 1.2 million new homes over the next five years, despite Treasury warnings this could not be achieved with present policy settings in place."

#auspol #nswpol #urbanism #UrbanPlanning #cities #AI #LLM #ChatGPT

Training experimental models to behave like a MAGA idiot, so that they learn how to suppress the worst 'persona' traits in LLMs.

Unfortunately you can't do this with people.
https://www.anthropic.com/research/persona-vectors

#IA #LLM

ils n'ont aucune limite et font toujours pire... dystopie de merde.

OpenAI (Chat-GPT) a pour projet de prendre le contrôle de l'enseignement supérieur en imposant ses outils à tous les recoins de la vie universitaire.

Les étudiants dépendront d'assistants IA, de l'intégration au diplôme. Bien sûr pour les cours, mais aussi les révisions, examens, et l'orientation professionnelle, les entretiens d'embauche. Tout.

https://www.nytimes.com/2025/06/07/technology/chatgpt-openai-colleges.html