"Here are the three inverse laws of robotics:

- Humans must not anthropomorphise AI systems.
- Humans must not blindly trust the output of AI systems.
- Humans must remain fully responsible and accountable for consequences arising from the use of AI systems."

Pretty decent for a start. https://susam.net/inverse-laws-of-robotics.html

#ai #llm

LLM (noun): a machine used by software teams to turn piles of cash into technical debt

#LLM #AI #software #TechDebt #TechnicalDebt #coding #AmbroseBierce

Where every major LLM stands politically

https://trakkr.ai/bias

#HackerNews #LLM #politics #AI #bias #tech #trends

So there is value in oral examination (literally oral).

At Stanford University:
«Some classes have started reintroducing proctoring - the supervision of candidates during an examination - and spoken-word tests to avoid cheating, [Lucy Zimmerman, a computer science major who served as a teaching assistant] said.»

This is, of course, a detail of a much bigger picture.

BBC article:
Stanford was their golden ticket - could AI help or hinder that?
<https://www.bbc.com/news/articles/c872j82j2qyo>

#AI
#ArtificialIntelligence
#Education
#LargeLanguageModels
#LLM
#LLMs
#UniversityEducation

Just reminding you all that #KeePassXC allows contributions from LLMs. https://keepassxc.org/blog/2025-11-09-about-keepassxcs-code-quality-control/ #KeePass #AI #LLM #Security #InfoSec

RE: https://mastodon.social/@radioscout/116795938482887745

WTF? AI companies are buying antique and second hand books from shops and markets in bulk to ship them to the US. They use them as further 'training fodder' for their LLMs on things they don't have, yet. Afterwards, they're dumped (for 'recycling'), thus depriving and depleting the (human) market to benefit from them. So they're on top creating scarcity to force people towards their own products destroying our planet and humanity.

Note: Article in German.

#antiquebooks #books #GenAI #LLM

I thank that so many people have engaged in the many threads & hundreds of posts re: #SFC's #LLM-gen-AI Recommendations¹ . The threads are a bii scattered in a diaspora around the #Fediverse now as per usual #Mastodon deletion situation, threads go wonky even if a few people delete I post. I delete-and-redrafted a few of mine to fix massive typos, but I think others were deleting. I can only find you if @-mention me at this point. However …

¹ https://sfconservancy.org/llm-gen-ai/llm-backed-generative-ai-recommendations.html

#OpenSource #LLMGenAI #AI

… Meanwhile, I've done a true marathon here to engage with the community. I'm going to “call it here” as being open 24/7 (& with a few people getting nasty at me when I didn't reply for a few days — literally accusing me of horrible behavior merely because I ran out of time on one thread and was answering another) has to end sometime.

My personal Q&A about #SFC's #LLM-gen-AI Recommendations is closed, but don't despair! I'll talk with my SFC colleagues about making a few weekly Q&A set times.

#UK Official Promises Statements 'Around VPNs' and Further Teen Restrictions on Chatbots and #SocialMedia

https://news.slashdot.org/story/26/06/21/2052213/uk-official-promises-statements-around-vpns-and-further-teen-restrictions-on-chatbots-and-social-media

#VPN #SocialmediaBan #AgeVerification #privacy #politics #LLM #AI

From comments: You know it's serious when Rick Beato talks about RAM and storage 😁
Actually, he's got a pretty sane point, the LLMs are being built for mass surveillance first, the rest is just byproduct.
https://www.youtube.com/watch?v=aXy8mQeuObk
#LLM #AI

From comments: You know it's serious when Rick Beato talks about RAM and storage 😁
Actually, he's got a pretty sane point, the LLMs are being building for mass surveillance first, the rest is just byproducts.
https://www.youtube.com/watch?v=aXy8mQeuObk
#LLM #AI

Training an LLM on a heavily cleaned, de-identified corpus can be like correcting every grammatical mistake in a large collection of texts: the result may look cleaner, but it can also lose the context, variation, and imperfections that reflect real-world language and behaviour.

A corpus scrubbed of every sensitive detail and irregularity can become a polished imitation of reality. Privacy protection could be necessary, but a model trained mostly on synthetic or over-sanitised data risks producing equally synthetic answers: coherent on the surface, yet disconnected from the messy context in which real life happens.

So what do you expect from an LLM following the regulation? and over-cleaning the input data?

#llm #ai

Ref: https://ora.ox.ac.uk/objects/uuid:fa1155e9-c2ff-436a-8391-455b622f4e64/files/r3b5919575 "AI models collapse when trained on recursively generated data"

Dovrei citare ogni episodio, ogni singola parola, congiunzioni comprese, perchè credo di non avere mai trovato qualcosa che non condivido in questo podcast capolavoro!

Grazie Walter, davvero!
Quando ti ascolto, mi sento un po' meno solo 🤗

"Come in questo caso, l'illecita raccolta di fotogrammi dai video di sicurezza, l'illecito c'è a prescindere.

Detto questo, facciamo un altro passo: non è che siccome c'è la videosorveglianza il primo pirla che passa se la guarda e raccoglie le figurine. Esiste un elenco di persone, ed è molto ristretto, che hanno la possibilità, in casi molto particolari, di accedere ai video di sorveglianza.

E nei posti governati da gente con un cervello, la procedura richiede sempre almeno due persone: tipicamente, un tecnico e un manager, per riesaminare i video. Quindi OK i tramvieri laidi, ma chi gli ha dato accesso ai video ha fatto ancora di peggio, perché gli incaricati della gestione dei video di sicurezza hanno gli stessi obblighi di riservatezza e confidenzialità degli amministratori di sistema, e infatti sono spesso le stesse persone.

Io direi che, oltre ai laidi, deve saltare pure la testa di qualcuno nell'area tecnica e qualche manager.

Certo, in Italia perfino gli amministratori di sistema delle banche si mettono a ficcare il naso nei conti correnti della gente, figuriamoci in un'azienda di trasporti pubblici.

Ma non è un problema. Le leggi esistono, si prende questa gente, si predono i loro superiori che hanno l'obbligo di sorveglianza, si prende la gente che ha fatto gli audit e garantito la compliance, si prendono quelli che hanno approvato un sistema di videosorveglianza dove gli accessi ai video non vengono loggati, o se vengono loggati non vengono auditati e li si accompagna tutti, gentilmente, alla porta.

A CALCI IN CULO.

Perché mi sono anche rotto il cazzo di questa privacy di carta dove per fare una foto della festa di Natale ufficio ci sono moduli in triplo originale, e poi il primo coglione con la password di amministratore fa il cazzo che gli pare."

DK 10x36 - SpaceX, ma anche i tramvieri laidi

#dataknightmarelalgoritmicoepolitico #dataknightmare_it #podcast #privacy #llm #aislop #tecnologia
 @DataKnightmare

#nowlistening con
@FraNiky

https://dk.dataknightmare.eu/spacex-ma-anche-i-tramvieri-laidi/

Just one more LLM, bro. Just one more LLM and this code will surely work fine!

https://github.com/osquery/osquery/issues/8939

#AI #LLM #NoAI #NoLLM #slop

#LLM agents violate Hammerstein's Law, clause 4:

"One must beware of anyone who is both stupid and hardworking; he must not be entrusted with any responsibility because he will always only cause damage."

https://en.wikiquote.org/wiki/Kurt_von_Hammerstein-Equord

#Firefox #AI #Chatbot feature exposed users to #email theft risk

https://cyberinsider.com/firefox-ai-chatbot-feature-exposed-users-to-email-theft-risk/

#privacy #cybersecurity #LLM #Mozilla

If you hit a writer's block, it's because you don't know something. Learn something!

And don't for the love of god try to fix it by using a fucking #LLM

#noAi #ai #art #writersblock

@bkuhn @conservancy @dalias

> While it was surely unintentional, your reply is quite similar to the cruelty of traditional #FreeSoftware rhetoric — wherein we shunned people for using #Apple & #Microsoft. Such users deserve sympathy and help toward more software freedom.
>
> Same goes for #LLM-backed generative #AI users.

Sorry, I usually really value your opinions on stuff, as you tend to run contrary to the so-pragmatic-they'd-pimp-their-own-mothers attitude that is far too prevalent in (F)OSS, but hard disagree.

There's a big difference between knocking someone for using a MacBook and knocking someone for using an LLM.

Using a MacBook harms the user in ways that are somewhat debatable. Its harm outside of the user is very negligible and even more debatable.

Using LLMs harms us all. I don't doubt that there are valid and ethical applications for LLMs in many sectors, but you don't build a giant baby shredder on top of an indigenous burial ground and then suddenly pivot to saying, "But wait, it can process potatoes, too!"

If we're not talking about the theft of human creativity, the theft of potable water, the theft of human labor and wages, the promotion of über-fascist plutocrats, and all of the other horrendous harms associated with LLMs, then we're not talking about LLMs. We're engaging in some kind of fantasy roleplay.

Respectfully,

—Some dingbat named Dane, or something, who loves humanity and freedom.

In a policy statement¹, @conservancy said:
> “#FOSS projects should not shun contributors who choose to use LLM-gen-AI systems.”

@dalias' reply:
>> “LOL WTF NOPE. 🤡”

While it was surely unintentional, your reply is quite similar to the cruelty of traditional #FreeSoftware rhetoric — wherein we shunned people for *using* #Apple & #Microsoft. Such users deserve sympathy and help toward more software freedom.

Same goes for #LLM-backed generative #AI users.

¹ https://sfconservancy.org/llm-gen-ai/llm-backed-generative-ai-recommendations.html

#OpenSource

When you're building software aimed at a niche related to hype, the only thing you can do is hook people onto it. I mean, you're targeting the kind people who have an attention span of a fruit fly. Unless they're literally addicted to your software, they're going to forget it as soon as they notice the next shiny thing.

#AI #LLM #NoAI #NoLLM #FreeSoftware #OpenSource

GLM 5.2 is really ripping it in benchmarks. For being roughly Opus-tier it's not as big as you'd think (~750B-A40B). All those bastards that shelled out for a 512GB mac studio can run it at home.

#llm

#OpenAI investigated by coalition of state attorneys general

https://www.msn.com/en-us/money/companies/openai-investigated-by-coalition-of-state-attorneys-general/ar-AA25wmMD

#privacy #AI #ChatGPT #LLM

Given these days you can't even expect #Gentoo contributors to be respectable, I'm working on adding a git hook that rejects commits with #LLM attribution. Could you help me find all the common patterns used to mark LLM-assisted #git commits?

So far I'm checking for author and Co-authored-by using the following e-mail patterns:

• copilot@github.com
• *@anthropic.com
• claude@users.noreply.github.com
• *+claude[bot]@users.noreply.github.com
• *@openai.com
• *+chatgpt-codex-connector[bot]@users.noreply.github.com
• *@cursor.com
• *@x.ai
• *@google.com

I think some people came up with some other tags to mark LLM commits but can't find that right now.

EDIT: added Assisted-by.

#NoAI #NoLLM #AI

What a good read!

> A growing economy used to be a rising tide that lifted all boats. Post-AI, it may only lift a handful of yachts.

It's really long, but very well worth it. Puts some real substance and lessons from history against arguments, to try and paint an extrapolated picture of what #BigAI will likely do to us and our society.

**A true risk that #AI poses to liberal democracies.**

https://matthewbutterick.com/extinction-level-capitalism.html

#GenAI #LLM #BigTech

LLMs don't "know" anything. They're just guessing the next word. 🤯

But that guess is built from 100B+ parameters, trained on basically the internet.

We explain how it works, why it lies sometimes, and how to run one (Gemma 4 / Qwen 3.6) on your own PC.
https://geekrealmhub.com/what-is-llm-guide-large-language-models-local-ai/

#AI #LLM #Tech