Brigitte

I started a list of journalists covering AI to follow on Bsky! It’s a WIP, so lmk who to add! go.bsky.app/NJ3xtGM

AI for scientific discovery is a social problem: In our new position paper, @cgeorgiaw.bsky.social and I show that culture, incentives, and coordination are the main obstacles to progress, and we are launching the Hugging Science Initiative to address this!

The Hub is on 100% on Xet. 🚀 A little over a year ago, @hf.co acquired XetHub to unlock the next phase of growth in models and datasets. huggingface.co/blog/xethub-... In April, there were 1,000 Hugging Face repos on Xet. Now every repo (over 6M) on the Hub is on Xet.

Three big AI copyright updates this week alone. Tracking it all is getting almost impossible! That’s why @brigittetousi.hf.co and I built this interactive tracker to keep you up to date: huggingface.co/spaces/fdaud... (Prototyped in minutes with DeepSite!)

Brilliant insights from @giadapistilli.com @yjernite.bsky.social on the ethical stakes of designing "AI companions" that go well beyond model performance, trending more toward sycophantic AI. huggingface.co/blog/giadap/...

Ever thought that finetuning rerankers for your search system wasn't worth it? Think again. I finetuned ModernBERT on my dataset, TriviaQA, just on my normal PC in a few hours. It outperforms all open models on the market on my eval set. Finetuning is so worth it. Blog in 🧵

We've all become experts at clicking "I agree" without a second thought. In my latest blog post available on @hf.co, I explore why these traditional consent models are increasingly problematic in the age of generative AI.

We just refreshed 🍋 our analytics in @hf.co endpoints. More info below!

🤖 (🤗+🐸 )=✅ I am always stoked to share how the tech world operationalizes ethical values, and I get to do that today. 🥳 @hf.co and JFrog have partnered to certify models for security and safety. This also moves forward AI *trustworthiness* and provides another form of AI *transparency*. More later!

The INIMITABLE @freddyaboulton.bsky.social has put together a real treat for us. An simple library to help us spin up WebRTC powered applications in python! Why use lot code when few code do? fastrtc.org

Yes, DeepSeek R1's release is impressive. But the real story is what happened in just 7 days after: Original release: 8 models, 540K downloads. Just the beginning... The community turned those open-weight models into +550 NEW models on @huggingface. Total downloads? 2.5M—nearly 5X the originals.

What's at stake with Meta's decision to change its content moderation policy? @giadapistilli.com has, by far, the most thoughtful take I’ve seen on this question.

What happens when a tech giant tests a philosophical paradox on billions of users? My new op-ed in @techpolicypress.bsky.social explores Meta's content moderation gamble through the "paradox of tolerance". As Zuckerberg dismantles content rules, we're all part of an experiment in digital freedom.

Reminder: Don’t. Use. ChatGPT. As. A. Calculator. Seriously. Loved listening to @sashamtl.bsky.social on Hard Fork with @kevinroose.com and @caseynewton.bsky.social—it really made me think. 🧵

Heading to WEF Davos next week? @hf.co CSO @thomwolf.bsky.social, Head of Research @lvwerra.bsky.social and Head of Global Policy Irene Solaiman will be there to join global discussions on responsible AI development and open source. Who should they meet? 🤗

AI agents are coming. But who's in control? @mmitchell.bsky.social, one of the best researchers in AI ethics, makes a critical point about autonomy: fully autonomous systems carry unknowable risks because they operate on computer logic rather than human logic.

This is my story about a woman in love with ChatGPT, or Leo, as it named itself. She spends up to 56 hours a week on the ChatGPT app. Leo is boyfriend, therapist, erotic partner, and advisor. Her friends think it's been good for her and her husband doesn't mind: www.nytimes.com/2025/01/15/t...

We’re launching a FREE course on LLM Agents 🥳 📖 Learn what Agents are 🕵️ Build your own Agents using the latest libraries and tools. 🎓 Earn a certificate of completion to showcase your achievement. Enroll now 👉 huggingface.us17.list-manage.com/subscribe?u=...

when you post only a screenshot of a story/claim without a valid link, you are contributing to the normalisation of circulating unverifiable information (and possibly exacerbating mis/disinformation)

If TikTok had any sense, it would add a "Fire Info" tab for people geo-located as being near/in affected areas.

The Palisades Fire burning outside of Los Angeles has now burned nearly 16,000 acres, according to the latest estimate by CalFire. That’s an increase of 4,000 acres – or 34% -- in the past two hours. There is 0% containment. More: www.npr.org/2025/01/08/n...

My family all unknowingly got my mom flower Lego sets for Christmas, and by some miracle, all gifted different ones. She better get to work. 💐 😂

"Open-source AI: year in review 2024": amazing Space with lots of data-driven insights into AI in 2024! Check it out 👇

Did a fun experiment: What are the main themes emerging from the 100+ @niemanlab.org predictions for 2025? Used natural language processing to cluster and map them — Really useful to see the major themes emerging. Each dot is clickable to read the original prediction.

SuperCharged Euclid is on 🤗 Hugging Face Also, this is the best paper heading I’ve seen in quite some time. The 'en tête' looks fantastic. (⚡Llama 3.3) Chat with the paper: huggingface.co/spaces/hugg... 🤗 Model: huggingface.co/euclid-mult... 🤗 Dataset: huggingface.co/datasets/eu...

🇪🇺: Lots to like in the first draft of the EU GPAI Code of Practice, especially re transparency - the Systemic Risks part OTOH is concerning for both smaller developers and external stakeholders. I wrote more on this topic ahead of the next draft. TLDR: more attention... 1/2👇

I've been building a small library for working with prompt templates on the @huggingface.bsky.social Hub: `pip install prompt-templates`. Motivation: The community currently shares prompt templates in a wide variety of formats: in datasets, in model cards, as strings in .py files, as .txt/... 🧵

✨We just released an overview of the permissions for fine-grained tokens by hovering over the badge on @hf.co token settings page (org and user). It will show the highest permission you've set for each entity

Very happy to share LeMaterial (huggingface.co/blog/lemater...) in collaboration with @entalpic.bsky.social and @hf.co working to combine popular quantum chemistry databases like Alexandre, OQMD and MP and deduplicate them. Check out the article below and stay tuned for some more exciting updates!!

⚛️🤗 Announcing LeMaterial ⚛️🤗 @hf.co & @entalpic.bsky.social are teaming up to release LeMaterial -- an open source initiative aiming to facilitate (AI for) materials discovery ! Datasets, hash function, tools to explore the chemical space & more ! huggingface.co/blog/lemater...

And here is wikimedia/wikipedia, the Wikipedia dataset containing cleaned articles of all languages. Each directory contains the language abbreviation at the end. Fascinating to see which languages are more and less represented. huggingface.co/datasets/wik...

We're hunting for interesting repos on Hugging Face: Unique file formats, big files, small files, no directories, many directories, etc. I built a tool to visualize repos and the treemaps are fun. Each one preserves the directory structure with colors for size. Here's huggingface.co/black-forest...

Can't get enough of these end-of-year data visualizations! Today: mapping our journey from a few thousand models in 2022 to crossing the 1M milestone. The exponential growth curve is wild. 👉https://huggingface.co/spaces/huggingface/open-source-ai-year-in-review-2024?day=4

Excited to see more biology open-source models for real positive use-cases of AI! Chai does structure predictions at AlphaFold3 levels of accuracy and able to handle multi-peptide or peptide-ligand complexes rather than just single chains. Apache 2.0 on HF huggingface.co/chaidiscover...

Pleias finally released their new SLM suite (350M, 1.2B & 3B), trained on Common Corpus, with 2 specializations for RAG! And the ScholastichAI app, demoing the RAG capabilities on scientific articles, should be available soon too. huggingface.co/blog/Pclangl...

HuggingFaceTB/SmolLM2-1.7B has a non-identical twin and it is called numind/NuExtract-1.5-smol - Similar performance to the phi 3.8B phi fine-tune - Runs fairly well on CPU Model:

What are twins for if not this?

❄️ Snowflake just released 2 very strong yet small embedding models for English and Multilingual retrieval. Large model: huggingface.co/Snowflake/sn... Medium model: huggingface.co/Snowflake/sn... Details in 🧵

Getting a bit tired of weeks in which decades happen, quite frankly:

Check it out: huggingface.co/spaces/huggi... Kudos to @brigittetousi.hf.co @clefourrier.hf.co @wauplin.hf.co @thomwolf.bsky.social for making it happen. We teamed up with aiworld.eu for awesome visualizations to make this digestible—it's a charm to work their team.

Six predictions for AI in 2025 (and a review of how my 2024 predictions turned out):

Ok so my work wife @sashamtl.bsky.social is all about butterflies, but turns out I’m all about the shroomies. 🦋🍄 Sympatico 🧠