I’m just going to play that South Park episode on a loop at people now. It was so spot on

This continues to be a recurring theme -_- Why is consumer agency so scary to the market? Noticed the same trend when I worked at Twitter: the personalization team cared for engagement metrics way more than actually showing users what they wanted to follow: “people don’t know what they like”

🤖 As AI-generated content is shared in movies/TV/across the web, there's one simple low-hanging fruit 🍇 to help know what's real: Visible watermarks. With others @hf.co, I've made sure it's trivially easy to add this disclosure to images, video, chatbot text. See how: huggingface.co/blog/waterma...

I wish there was a way for arxiv to take me to the abstract page by default instead of the html page 😬 it is strictly worse

I adore the public library. I just put 15 picture books on hold for my kiddo, on trains (current obsession) and beaches (first visit soon). We'll pick them up after they're all gathered at our branch. This is free! You all know this, I just feel the need to say it out loud occasionally.

It's interesting to think that the visual depiction of the word "AI" is still strongly coupled to humanoid robots even as the most popular applications of what we currently think of AI are not that. Terminology overloading over time is fascinating to live through.

The Kant Car

Research posts on Bluesky are more original — and get better engagement www.nature.com/articles/d41...

It’s fascinating to see the time gap between Taylor swift’s engagement insta post and full length news articles pop up. Journalist folks please tell me do you have these pre-written just in case or are you just that proficient?

Thinking a lot about documentation standards today. I would be very interested in people's thoughts on what factors lead to a documentation schema to become a standard? Thinking about datasheets vs model cards for eg, and how one became a (somewhat) durable standard..

My favorite thing about the Gemini chain of thoughts is how it says “I am now laser focused on” <insert part of code that I asked to fix>. It’s cute

I genuinely want to know the thought process here. Is each model iteration a new being? Is Claude 4.1 its own legal entity deserving of model welfare different from 4.0? Or is it like one human updating their world knowledge and becoming smarter? Was the very first trained Claude the robot embryo?

Remember the discussions around AI Eval plots last week? We at @eval-eval.bsky.social have a quick blog post on the AI Eval Charts Crisis!

The product decision to discontinue older versions of ChatGPT and the comments on Reddit around that decision reminded me once again of discussions around “robot death”, which is real insofar as people’s feelings and emotions are real.

New paper idea: Graph Crimes in Evaluations as a form of Epistemic Injustice

I’ve made my husband laugh with the same joke 3 times I feel so powerful

This! Also begets the question, why are people starting to get averse to social forms of knowledge production, involving multi-level stakeholders? Why is this mad rush to clip out the humans from the process, instead of working in tandem?

🚨 AI Evals Crisis: Officially kicking off the Eval Science Workstream 🚨 We’re building a shared scientific foundation for evaluating AI systems, one that’s rigorous, open, and grounded in real-world & cross-disciplinary best practices👇 (1/2) Read our new blog post: tinyurl.com/evalevalai

[New] Husbandposting! And yes we had a croquembouche for dessert because I saw it on masterchef once and I’ve always wanted that ❤️

New blog post alert! 🚨"What is the Hugging Face Community Building?", with @yjernite.bsky.social and Irene Soliaman The AI narrative focuses on big players, but the real story is happening in the open source AI ecosystem across 1.8M models, 450K datasets, and 560K apps, on @hf.co.

Gotta plug @timnitgebru.bsky.social @alexhanna.bsky.social and Asmelash Teka Hadgu's piece: www.dair-institute.org/blog/elders/ And @anikacolliernavaroli.com's excellent interview with Timnit on the subject: www.techpolicy.press/through-to-t...

Random hot take: Too many people are building AI for genz and not enough are building for their grandparents. Where are the apps/bots that say a retired, perhaps non English speaking grandma can use to improve their lives?

Not enough people know about nairrpilot.org and it shows. NAIRR needs better marketing and more resources to become a strong public computing infrastructure for American researchers and builders! 🇺🇸 @hf.co is one of many resources you can access through NAIRR ;)

I still think the set design of the Evita revival at the American Rep theater at Harvard was the most stunning interpretation of all time - I hope this concept makes it to Broadway at some point 🤩

Help this project out by participating if it’s up your alley! Moral KG is an ambitious attempt to map AI reasoning traces onto a developing massive philosophical ontology, and very much being built as a bootstrap effort with a rag tag team 🤗

Go say hi to @jennmickel.bsky.social and others and get to know what we are doing!

We're doing better AI Evals together! A herculean effort, but evals are in crisis and together we can go back to research roots and good science 🤗

Generative AI often renders the user invisible in their limited worldview. Please sign up for a short interactive workshop on AI, Misrepresentation and Mental Health, at both @facct.bsky.social in Athens, and Alt-FAccT in NYC! Limited space, so hurry! Sign up here! tinyurl.com/ai-mirrors

Alright, people, let's be honest: GenAI systems are everywhere, and figuring out whether they're any good is a total mess. Should we use them? Where? How? Do they need a total overhaul? (1/6)

I'm very pleased that both FAccT and AIES have way more qualitative evals of systems than I have seen in previous years. More of these please!

Living downtown and literally 2 blocks from the Opera House is certainly a clutch because I’m always late to things. Catch Roméo et Juliette playing in Boston it was great 😍

Can you train a performant language model using only openly licensed text? We are thrilled to announce the Common Pile v0.1, an 8TB dataset of openly licensed and public domain text. We train 7B models for 1T and 2T tokens and match the performance similar models like LLaMA 1 & 2

Competitive language models trained on fully openly licensed data is possible! Don’t let for profit orgs tell you otherwise.😉 Incredible work by some of my favorite folks across organizations.

I’ll be at alt-FAccT in New York on June 20! Skipping Athens this year for personal reasons (but a really good one 💍) so I’d love to see you if you’re going! 🤗

And now add AI Agents to the conversation, and we’re looking at a dire need for infrastructure overhaul

I miss the research hangout atmosphere of the old Twitter, it’s just all politics on X and on Bsky now - just on opposite sides of the spectrum. Sorry to say but it seems like special interest subreddits are the last bastion standing 😅

In the wake of the US-UAE data center announcement in Abu Dhabi I once again wish for stronger commitments to public infrastructure - could be as simple as letting a small fraction of new compute projects reserved for small open source developers to use!

Please help spread the word! Attending @facct.bsky.social 25 and want to host an in‑person or virtual social? Submit a short application to our Call for Socials to receive organizer support! facctconference.org/2025/cfsocial We welcome socials based on both research topics and personal identities✨

It’s not just biased—generative AI is importing old stereotypes into new languages and cultures. My latest for @wired.com: www.wired.com/story/ai-bia...

We're working on an announcement so keep an eye out for details but I can't WAIT to see how our accepted FAccT CRAFT workshop turns out. Yet another personal trauma --> shitpost on social media --> getting new collaborator DMs --> project pipeline!

It has been a year. And quite a year! The piece continues to be controversial, relevant and educational, and we’re looking for an in-person home for it! 👀

Thrilled to share that our paper: "It's not a representation of me": Examining Accent Bias and Digital Exclusion in Synthetic AI Voice Services - has been accepted at @facct.bsky.social 2025! - with @shiramichel.bsky.social , Sufi Kaur, Sarah Gilespie, Jeffrey Gleason and Dr. Christo Wilson.

🚨 New Article: Empowering Public Organizations: Preparing Your Data for the AI Era, with @yjernite.bsky.social Let’s discuss how public organizations can unlock the full potential of their data in the age of AI.

What are 🤖 AI agents, and what’re the ethics in developing them? New quick+easy read from me, @evijit.io , @sashamtl.bsky.social and @giadapistilli.com , in @technologyreview.com . 🤩 www.technologyreview.com/2025/03/24/1... (1/n)

On March 14, Hugging Face (@hf.co) submitted our response to the White House Office of Science and Technology Policy's request for information on the AI Action Plan. Here are the highlights:

Amazing work from HF colleagues that's super relevant to AI policy - open development matches "frontier" performance 🤗 For policymakers, it means we need regulation that works for both open and closed model - not closed-first that disproportionately harms open! Good news is... 1/2

H/T Anjali Singh for sending this pic. Every single day our paper seems more and more like a warning sign ;-; Automating away human oversight via cognitive laziness is in no one's best interests (other than maybe the model developer -- and even then what about liabilities?)

👩‍💻 What is AGI...and should it be a goal? 😱 My podcast interview on this question is out! With @eryk.bsky.social and @borhane.bsky.social , interviewed by @justinhendrix.bsky.social for @techpolicypress.bsky.social : www.techpolicy.press/should-agi-r... Based on this paper: arxiv.org/abs/2502.03689

I wonder if this mad rush to agentifying workflows is the precursor to pushing over liability to the system instead of the human operator: if the “agent”, with some level of agency, makes a poor decision, it’s the agent’s fault, right?

AI real talk. We (humanity) are moving full speed ahead at building AI agents for war that can create a runaway missile crisis of mutually assured destruction globally. Is the option of not allowing AI agents to deploy missiles already off the table, or is that still up for discussion?