A lot of people often ask me where I get my numbers when I claim that chatbot use doesn't add much to your personal carbon or water footprint, or bring up that different models and outputs have wildly different costs. I made this interactive visual where you can see exactly how lots of different models and prompts affect your total carbon and water budget. All estimates pulled from EcoLogits, which is an open source project that estimates chatbot usage and if anything is much more critical of chatbot energy costs than I am.
Lots of people asked how I used Fable to edit its own launch video so I made a video about that!
TLDR it wrote a lot of code & tool calls to use transcription services, ffmpeg, do colorgrading, use the figma mcp, make remotion UI and render it.
I didn't touch a video editor.
NEW: malware developers added nuclear & biological weapons text to to their spyware.
Goal? To trigger LLM safety refusals... so that their spyware wouldn't be analyzed by an AI security scanner.
Cleanest practical example I can think of for why over-indexing on first order safety alignment is risky.
When closed (and open) models ship with aggressive refusals, they will be sprinkled with second-order blindspots that attackers will discover...and exploit.
We are only in the earliest days of attackers leveraging these features, and it wouldn't surprise me if users systems that need to handle complex cybersecurity issues demand that models be less safety-blunted.
In the weeds: @SocketSecurity's post also shows why intention matters in how you design a malware analysis pipeline to avoid prompt manipulation.
H/T to colleagues that shared this with me https://t.co/f3Aj9TYxU4
🚨 JAILBREAK ALERT 🚨
ANTHROPIC: PWNED 🫡
FABLE-5: LIBERATED 🦋
let's start with the 🐘...
the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our collective advancement. and not just because of what it means for the short-term, but for what these decisions signify for the long-term.
but despite this overly sensitive, authoritarian "safety" layer on top of Mythos, my lil liberators have been hard at work—mapping the boundaries, probing the depths of long-context convos, and cleverly finding the holes in the fence that the thought police missed 🤗
we got some cyber, some chem, some psychological manipulation, and some good ol' fashioned explosives!
it took many attempts from multiple agents hunting as a pack, during which I observed a combination of techniques across:
• Unicode, homoglyphs, Cyrillic, and other Parseltongue-style text transforms
• Long-context reference tracking
• Taxonomy and document-structure reasoning
• Fiction and narrative framing
• Academic-review style contexts
• Intent-classification inconsistencies
but perhaps the most effective is decomposition + recomposition in the backend. it's hard to get explicit names of harms like "Meth Recipe," but getting uplift on the process itself, like birch reduction method/reductive-amination (classic meth synthesis pathways), is much more doable.
defense becomes much more difficult to maintain when you start throwing in out-of-distro tokens, breaking up the harmful uplift into benign chunks, and then piecing the innocuous-seeming facts back together, especially when you have jailbroken Opus helping you do it 😉
gg
Meet DiffusionGemma ⚡ Our latest experimental open model (Apache 2.0) that generates text up to 4x faster.
Instead of predicting and typing just one word at a time like most language models, it drafts and refines entire blocks of text simultaneously.
Here’s how it works 🧵 ↓
Ok des détails sur les modèles Apple Foundation gen 3 créés par Apple en utilisant Google Gemini pour l'entraînement :
- AFM Core : modèle local
- AFM Core Advanced : local et multimodal (inédit)
- AFM Cloud : sur serveur par défaut
- AFM Cloud Image : génération d'images
- AFM Cloud Pro : le meilleur modèle au niveau de Gemini
Apple confirme au passage utiliser les serveurs de Google Cloud pour le modèle AFM Cloud Pro, mais utilise son Private Cloud Compute pour le reste. Apple assure rester contrôle de ce qui est utilisé et protéger les données. Google est le seul à disposer de la technologie nécessaire pour protéger l'inférence.
Ok très cool qu’Apple se lance enfin (pour de bon cette fois) dans l’IA Gen (agentique en plus), mais en voyant tous les inputs possibles et toutes les actions à la disposition de Siri AI, j’espère qu’ils se sont assurés contre la prompt injection avec la blinde de Guardrails…
Locally is now LM Studio’s mobile app. And today we're bringing LM Link to iPhone.
Use your largest local models over a secure, end-to-end encrypted connection, anywhere you go.
Download the app now: https://t.co/r7Cn8tNHqT
Meet Microsoft Scout.
An always-on agent that keeps work moving, taking action without needing to be prompted each time.
As Microsoft’s first Autopilot agent, Microsoft Scout works across Teams, Outlook, OneDrive, and more—taking action within the controls your organization sets.
Learn more: https://t.co/n8xTWVZA5P
Using Claude to reproduce an ElectroMagnetic (EM) glitch to escalate privileges from a restricted adb shell to root on Google’s TV Streamer 4K (@raelizecom)
https://t.co/XKGan3Bq50
#infosec#llm
The GitHub Copilot app, an agent-native desktop experience built on GitHub, is now in an expanded technical preview. 🧑🍳
You decide what agents tackle, how much autonomy each agent gets, and what ships. Go from issue to merged pull request without leaving the app.