Margaret Mitchell

@mmitchell.bsky.social

3293

Followers

302

Following

980

Posts

Interdisciplinary researcher focused on shaping AI towards long-term positive goals. ML & Ethics.

Posts Replies Media

Another ethics-informed bit of norm-setting in AI we've started @huggingface.bsky.social : Providing metadata about the private information in datasets. huggingface.co/blog/presidi... Applicable to: Dataset Cards, Auditing, impact assessments, data measurement

I've played but a smol role in this, and it's been amazing to work with so many people I respect and admire. Check out the consensus we've ended up with on evaluating social impact! arxiv.org/abs/2306.05949

Evaluating the Social Impact of Generative AI Systems in Systems...arxiv.org Generative AI systems across modalities, ranging from text (including code), image, audio, and video, have broad social impacts, but there is no official standard for means of evaluating those...

📣 Pleased to share the newest newsletter from the Ethics & Society regulars @huggingface.bsky.social ! We dig into one of the pressing issues in AI at the moment: 📜DATA QUALITY! Led by Lucie-Aimée and @evijit.io . 🤗 huggingface.co/blog/ethics-...

Ethics and Society Newsletter #6: Building Better AI: The Importance of Data Qualityhuggingface.co We’re on a journey to advance and democratize artificial intelligence through open source and open science.

Reposted by

Margaret Mitchell

excellent investigation, beautiful visualisation, and scary findings. tech corps are burning our planet down so people can stop making art www.bloomberg.com/graphics/202...

AI Is Wreaking Havoc on Global Power Systemswww.bloomberg.com New artificial intelligence data centers are coming online so fast that the electricity demand is straining global power grids and threatening clean energy goals.

Yesterday my much-loved aunt Maggie passed away. She was a force of nature, brilliant and thoughtful. A Hollywood metal artist who defined an aesthetic you're probably familiar with. Sharing a little video about her work on Star Trek & beyond. www.youtube.com/watch?v=Gqqk...

Maggie Schpak: STAR TREK Metalwork Artistwww.youtube.com Industrial designer Ryan Norbauer and jewelry historian Karen Christians visit the workshop of Maggie Schpak who has created some of the most iconic objects ...

Reposted by

Margaret Mitchell

King's College London is hiring a Lecturer (Assistant Prof.) with a focus on authoritarianism, populism, democracy, human rights, political communication, transitional justice. #polisky #cpesky #PSJMInfo

Lecturer in International Development (Politics)www.kcl.ac.uk

Did you know: there are limited solutions for curbing Deepfake Porn? We must foster more informed friend groups, better platform processes, legal recourse & technical solutions to help those most harmed by this -- teen girls. New from me in @teenvogue.bsky.social www.teenvogue.com/story/how-to...

How to Stop Deepfake Porn Using AIwww.teenvogue.com AI experts explore the tools working to make deepfake porn harder to make.

"I want AI to do my laundry and dishes so that I can do art and writing", not the other way around. @huggingface.bsky.social : OK! venturebeat.com/ai/hugging-f...

Hugging Face and Pollen Robotics show off first project: an open source robot that does choresventurebeat.com Pollen Robotics partnered with Hugging Face to train the robot to do a variety of household chores and safely interact with humans and dogs.

Why do me and some of me awesome woman/NB colleagues @huggingface.bsky.social -- @sashamtl.bsky.social Liz Allendorf @witko.bsky.social Bruna Trevelin -- suddenly have bios in @teenvogue.bsky.social ? 🫢 Answer coming soon... www.teenvogue.com/contributor/...

Margaret Mitchell, Sasha Luccioni, Elizabeth Allendorf, Emily Witko, Bruna Trevelinwww.teenvogue.com Margaret Mitchell is chief ethics scientist at Hugging Face, Sasha Luccioni is the company's AI research scientist, Elizabeth Allendorf is the backend engineer, Emily Witko is head of culture and DEIB...

Why do me and some of me awesome woman/NB colleagues @huggingface.bsky.social -- @sashamtl.bsky.social Liz Allendorf @witko.bsky.social Brun Trevelin -- suddenly have bios in @teenvogue.bsky.social ? 🫢 Answer coming soon... www.teenvogue.com/contributor/...

Like a breath of fresh air to experience so many brilliant minds engaging in deconstructing groupthink in tech. #Facct2024

I no longer care as much when papers don't cite my work when the authors were not only aware of it, but I had advised them. So that's nice. Tech can be rough, as you get older, some of the apathy that sets in can be quite pleasant.

Reposted by

Margaret Mitchell

The Pillars of a Rights-Based Approach to AI Development - @mmitchell.bsky.social: techpolicy.press/the-pillars-...

The Pillars of a Rights-Based Approach to AI Development | TechPolicy.Presstechpolicy.press By centering peoples' rights we can shape AI to maximally benefit humanity while minimizing harms, says Margaret Mitchell.

PSA: I'm recently obsessed with datatrove for processing large datasets. 10/10 would recommend for all your Large NLP data needs. github.com/huggingface/...

GitHub - huggingface/datatrove: Freeing data processing from scripting madness by providing a set of platform-agnostic customizable pipeline processing blocks.github.com Freeing data processing from scripting madness by providing a set of platform-agnostic customizable pipeline processing blocks. - huggingface/datatrove

AI and human rights: 'Domicide' refers to the systematic and deliberate mass destruction of housing during violent conflict. The UN has called out how AI is being used specifically for destroying this human right in Gaza. 🧵 www.ohchr.org/en/press-rel...

Gaza: UN experts deplore use of purported AI to commit ‘domicide’ in Gaza, call for reparative approach to rebuildingwww.ohchr.org GENEVA (15 April 2024) – UN experts today deplored the purported use of artificial intelligence (AI) and related military directives by Israel in occupied Gaza leading to an unprecedented toll on the ...

Prediction: Lasting impact of the "screw you effect". This is when a participant deliberately provides false info to influence a study's outcomes. We will see this more now when people share "facts" online, as these "facts" are being nonconsensually consumed for AI knowledge.

Seriously, tho. As kooky as the term "Google" is, to "Google" something has, until now, meant to retrieve high quality information about it. This fundamental has shaped society and cemented Google's dominant marketshare. What’s happening now really rides home how G has changed.

Google's AI-generated summary of how AI works is a bit of a mess. My favorite part is how the AI selects its own data before being trained, that's a neat trick.

Reposted by

Margaret Mitchell

I didn’t know half this stuff

Critical reading on LLM eval from my amazing coworker @clefourrier.bsky.social. huggingface.co/blog/clefour...

Let's talk about LLM evaluationhuggingface.co A Blog post by Clémentine Fourrier on Hugging Face

It's heartening to see people from all different backgrounds opining intelligently on the ethics of AI and consent today. Sort of a rare moment worth pausing to appreciate.

This is a great piece connecting some of the modern approaches to AI with the concept of "manifest destiny" (public be damned). Plainly unpacks the sense of nausea and unease many of us have. 10/10, recommended reading.

wrote about open ai and manifest destiny www.theatlantic.com/technology/a...

The final version of the approved EU AI Act even specifically recognizes model cards and datasheets: The work that me and my former Google Ethical AI team co-lead Timnit Gebru spearheaded.

Happy "the EU AI Act has been approved" day! Has so much that people concerned about social impact of AI have been working towards. I'm thankful and impressed to see the ideas put into law. www.consilium.europa.eu/de/press/pre...

Artificial intelligence (AI) act: Council gives final green light to the first worldwide rules on AIwww.consilium.europa.eu Council gives final green light to artificial intelligence act, the first law of such kind in the world.

Tickets to FAccT purchased. Come join us in Rio to learn about the newest research/tools for ethical AI, and AI critiques that center those directly affected. Plug for my tutorial on building Ethical Charters, with @giada.bsky.social, who is a for-real AI ethicist.

I was invited to guest host a MAIHT3000! Tune in noon PDT, May 20th, for a discussion with me and @emilymbender.bsky.social on facts & fantasy underlying the idea of "AI deception". twitch.tv/dair_institute

On the next Mystery AI Hype Theater 3000, guest host @mmitchell.bsky.social and I will look into claims that "AI has learned to deceive humans". Join us live at noon Pacific, May 20, 2024: twitch.tv/dair_institute

dair_institute - Twitchtwitch.tv Twitch account for The Distributed AI Research Institute (DAIR).

Reposted by

Margaret Mitchell

From the archives: www.theonion.com/these-kids-s...

Reposted by

Margaret Mitchell

Begging for just one corner of life not to be seen as an appropriate next site for an AI surveillance tool.

Okay, I'm gonna stop doing things for other people and code now and no one can stop me.

By the way, if you want to analyze almost any current SOTA LLM, @huggingface.bsky.social has a thing where you can super easily spin it up for a few bucks/hr, mess with it, then shut it down: ui.endpoints.huggingface.co I'm kind of addicted.

Inference Endpoints - Hugging Faceui.endpoints.huggingface.co Deploy any Hugging Face model in minutes.