veganism.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
Veganism Social is a welcoming space on the internet for vegans to connect and engage with the broader decentralized social media community.

Administered by:

Server stats:

293
active users

#LLMs

74 posts60 participants2 posts today

🔐 LLMs are missing 40%+ of cyber attack data...
...while reporting their answers with sky-high confidence.
Dangerous combo, right?

The latest study shows:
Confidence ≠ Accuracy when it comes to AI in cybersecurity.

So what happens when your model confidently flags the wrong vulnerability?

💬 Have you seen an AI "hallucinate" in your security workflows?

👇 Join the conversation + read more here:
blueheadline.com/cybersecurity

LLMs Are Dangerously Confident When They’re Wrong in Cybersecurity - Blue Headline Tech
Blue Headline · LLMs Are Dangerously Confident When Wrong In CybersecurityLLMs are overconfident and inconsistent in cybersecurity tasks, often making critical CTI mistakes with high certainty. Here’s why that’s a problem.

@Legit_Spaghetti

I don't know what you are complaining about.

Looking at the image: The single elephant and beach are in it, as requested; and the beach is definitely not within the elephant, but indeed is without.

Yes, I have indeed experienced Alexander's hymn "There is a green hill far away". Many, many times.

Did you try "sans"?

You could try to see what it makes of the first two lines of the hymn.

(-:

#AI#LLMs#elephants

This tells us a lot about how the lives of an increasing number of human beings are so empty of social contact with other human beings that they need to enter into false relationships with chatbots governed by neural networks and statistical probabilities...

"More and more of us are using LLMs to find purpose and improve ourselves.

Therapy and Companionship is now the #1 use case. This use case refers to two distinct but related use cases. Therapy involves structured support and guidance to process psychological challenges, while companionship encompasses ongoing social and emotional connection, sometimes with a romantic dimension. I grouped these together last year and this year because both fulfill a fundamental human need for emotional connection and support.

Many posters talked about how therapy with an AI model was helping them process grief or trauma. Three advantages to AI-based therapy came across clearly: It’s available 24/7, it’s relatively inexpensive (even free to use in some cases), and it comes without the prospect of judgment from another human being. The AI-as-therapy phenomenon has also been noticed in China. And although the debate about the full potential of computerized therapy is ongoing, recent research offers a reassuring perspective—that AI-delivered therapeutic interventions have reached a level of sophistication such that they’re indistinguishable from human-written therapeutic responses.

A growing number of professional services are now being partially delivered by generative AI—from therapy and medical advice to legal counsel, tax guidance, and software development."

hbr.org/2025/04/how-people-are

Harvard Business Review · How People Are Really Using Gen AI in 2025Last year, HBR published a piece on how people are using gen AI. Much has happened over the past 12 months. We now have Custom GPTs—AI tailored for narrower sets of requirements. New kids are on the block, such as DeepSeek and Grok, providing more competition and choice. Millions of ears pricked up as Google debuted their podcast generator, NotebookLM. OpenAI launched many new models (now along with the promise to consolidate them all into one unified interface). Chain-of-thought reasoning, whereby AI sacrifices speed for depth and better answers, came into play. Voice commands now enable more and different interactions, for example, to allow us to use gen AI while driving. And costs have substantially reduced with access broadened over the past twelve hectic months. With all of these changes, we’ve decided to do an updated version of the article based on data from the past year. Here’s what the data shows about how people are using gen AI now.

"If you’re new to prompt injection attacks the very short version is this: what happens if someone emails my LLM-driven assistant (or “agent” if you like) and tells it to forward all of my emails to a third party?
(...)
The original sin of LLMs that makes them vulnerable to this is when trusted prompts from the user and untrusted text from emails/web pages/etc are concatenated together into the same token stream. I called it “prompt injection” because it’s the same anti-pattern as SQL injection.

Sadly, there is no known reliable way to have an LLM follow instructions in one category of text while safely applying those instructions to another category of text.

That’s where CaMeL comes in.

The new DeepMind paper introduces a system called CaMeL (short for CApabilities for MachinE Learning). The goal of CaMeL is to safely take a prompt like “Send Bob the document he requested in our last meeting” and execute it, taking into account the risk that there might be malicious instructions somewhere in the context that attempt to over-ride the user’s intent.

It works by taking a command from a user, converting that into a sequence of steps in a Python-like programming language, then checking the inputs and outputs of each step to make absolutely sure the data involved is only being passed on to the right places."

simonwillison.net/2025/Apr/11/

Simon Willison’s WeblogCaMeL offers a promising new direction for mitigating prompt injection attacksIn the two and a half years that we’ve been talking about prompt injection attacks I’ve seen alarmingly little progress towards a robust solution. The new paper Defeating Prompt Injections …

"Finally, AI can fact-check itself. One large language model-based chatbot can now trace its outputs to the exact original data sources that informed them.

Developed by the Allen Institute for Artificial Intelligence (Ai2), OLMoTrace, a new feature in the Ai2 Playground, pinpoints data sources behind text responses from any model in the OLMo (Open Language Model) project.

OLMoTrace identifies the exact pre-training document behind a response — including full, direct quote matches. It also provides source links. To do so, the underlying technology uses a process called “exact-match search” or “string matching.”

“We introduced OLMoTrace to help people understand why LLMs say the things they do from the lens of their training data,” Jiacheng Liu, a University of Washington Ph.D. candidate and Ai2 researcher, told The New Stack.

“By showing that a lot of things generated by LLMs are traceable back to their training data, we are opening up the black boxes of how LLMs work, increasing transparency and our trust in them,” he added.

To date, no other chatbot on the market provides the ability to trace a model’s response back to specific sources used within its training data. This makes the news a big stride for AI visibility and transparency."

thenewstack.io/llms-can-now-tr

The New Stack · Breakthrough: LLM Traces Outputs to Specific Training DataAi2’s OLMoTrace uses string matching to reveal the exact sources behind chatbot responses

From #DeepSeek to #Qwen: discover how to port and optimize large language models (#LLMs) on Rockchip #RK3588-C #SBC and NVIDIA Jetson Xavier NX #AI Box! This guide covers:
✅ NPU Deployment using RKLLM-Toolkit for high efficiency
✅ Ollama for CPU-Based Deployment – quick & flexible
✅ Enhanced Interactions with Chatbox UI & Web UI
Explore AI model porting & interactive solutions for embedded AI development!👇
forlinx.net/article_view_675.h

www.forlinx.netFrom DeepSeek to Qwen: Expert Guide on AI Model Porting & Interaction - Blog - Forlinx Embedded Technology Co., Ltd.Expert guide on porting AI large models from DeepSeek-R1 to Qwen, covering NPU/CPU deployments and interactive UI solutions like Chatbox and Web UI. Contact us.

Commercial #LLMs are shifting right (links in screenshot alt text). This value shift is not a coincidence, but done intentionally by the corporations behind them (#OpenAI, #Meta...).

This is a extremely serious problem. People increasingly use genAI as their sources for "truth" or facts, even for mundane inquiries.

With enough time and interactions, this COULD BE a way for #AI to use a latent "onboarding program" where users are increasingly exposed to (alt-) right adjacent ideas.

A solution for now might be to use fully open LLMs (#Olmo2 is one of the few) and to making transparancy tools like transluce.org mandatory for AI corporations.

BUT it is important for schools, universities and others in #education to refrain from using AI systems from companies doing this. (Looking at #fobizz, #bwgpt and so on).

We should stop focusing on "skills" and "competencies" when it comes to AI, but instead ask for sovereignty - KI-Mündigkeit.

🚀🎮 Oh joy, the existential crisis of running #NixOS #WSL just to keep those all-important #LLMs happy on your #gaming rig. Because who doesn't love turning their PC into a never-ending experiment in software masochism? 🤖💻
yomaq.github.io/posts/nvidia-o #SoftwareExperimentation #TechHumor #HackerNews #ngated

yomaq · Nvidia on Nixos WSL - Ollama up 24/7 on your gaming PCConvenient LLMs at Home