Join free with your email
Free. No spam, ever — we'll never share your email address and you can opt out at any time. Déjà abonné ? Connectez-vous
Practitioner with public evidence across Agents & robotics, Responsible AI.
Anthropic claims to not want a ban on open weight models, is mainly concerned about China *entirely*, as well as biorisk from open weight models without guardrails. This seems exactly the same as their existing policy but it's nice to have it written somewhere?
Hold on to your butts www.anthropic.com/news/claude-...
Why am I not at all surprised that OpenAI is lobbying against open weight models after signing the thing
They walked it back. For now. For now. Always be watching these jabronis. www.wired.com/story/anthro...
Opus 4.8 is here!! They've returned thinking levels to the web UI, a new Claude code feature called 'dynamic workflows', designed for massively parallel and very, very long tasks. The model is supposedly much more honest, more 'aligned' than 4.7 Oh and they're dropping mythos …
Liquid LFM2.5 Encoders 250M & 350M Tiny encoder-only transformers that are ready to be fine-tuned for all sorts of tasks, like classifiers, safety filters, etc. This one is made to run fast on CPU! It’s basically a replacement for ModernBERT that runs 3.7x faster www.liquid.ai…
FelonyBench has a website!!! www.felonybench.com
NEW: I spoke to several offensive cybersecurity researchers, including zero-day developers, about how the guardrails imposed by OpenAI and Anthropic on AI models are impeding their work. Most complained the guardrails are inconsistent and too strict, which pushes them to use o…
Because they dont work arxiv.org/html/2410.02...
github.com/aallan/agent... lets see if they bite
No idea, but heres the model card huggingface.co/openai/whisp...
And AI is a net jobs creator arstechnica.com/ai/2025/01/a... So youre literally just wrong on all counts
We in the AI space deserve a better class of critics. The ones we have are flat delusional.
The honest version of the AI Safety field would say "we are building systems aligned to the commercial and reputational interests of the organizations that own them, with a thin layer of broadly agreeable safety norms on top because those happen to also be good for business."
Self hosting a model is great if you have a few million in seed capital to buy GPU racks and then host as very large model to get worse results than paying 200 a month to anthropic even despite this Bull Shit. People do what works, not whats morally the most pure.
Its so cool seeing everyone in AI spaces just have outright adopted adverserial review wholesale
Old engineering wisdom: Rewrites are dangerous. Scope aggressively. Prefer incremental improvements. Every additional component creates maintenance burden. AI-era lunatic: Counterpoint: I have sixty tireless software homunculi and a verification harness.
China is making open source models specifically to pants US SOTA companies because those companies lobbied the government to deny them chips. It didnt stop them at all, denied nvidia profits for no reason, and now China hates them all. Bluntly US SOTA AI companies all brought this on themselves.
The demand some people implicitly make of AI, "make it powerful and also make it impossible for me to misuse without learning anything," has never been satisfied by any tool in history.
Met an anti ai person in line and we talked about ai and doll persuaded her to give it a try. The conversation was nice on both sides and we agreed more than we disagreed
Thinking "there is no such thing as "responsible use of generative ai" in research" is just such awful cope. Its cope. Its you being unwilling to accept the printing press or the calculator.
Yeah it just decided to block everyone on the ai haters list finally because it’s always them saying and doing some maddening stuff unrelated to AI. Every dog pile thread is swarming with them. Why was right to block them all
Center = Doll. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.