Thread by @math_rachel

I have long admired @timnitGebru for her brilliance, moral courage, clear voice in speaking up for what is right, & influential scholarship. It is truly terrible that Google would do this.

In this thread, I want to share some of Timnit's work I love

https://twitter.com/timnitGebru/status/1334352694664957952?s=20

I've quoted "Datasheets for Datasets" (2018) in many of my talks & assign it as reading in my class. It highlights decisions that go into creating & maintaining datasets, and how standardization & regulation came to other industries

arxiv.org/abs/1803.09010

Timnit worked with @jovialjoy on the original GenderShades research, which has had a profound impact on facial recognition, led to concrete regulations, and changed our industry

https://twitter.com/math_rachel/status/976235105973714949?s=20

It is rare for academic work to have this big of a practical impact:

https://twitter.com/math_rachel/status/1163637485533929473?s=20

"Lessons from Archives" is a great paper by @unsojo & @timnitGebru on what machine learning can learn from the library sciences about data collection, in light of as consent, power, inclusivity, transparency, and ethics & privacy:

https://twitter.com/math_rachel/status/1223799130180349953?s=20

I love this passage from @timnitGebru in NYT on what a narrow framing of bias as just error rates across groups misses: whether a task should exist at all, who deploys it, and on which population,...

https://twitter.com/math_rachel/status/1275515335492304896?s=20

Timnit is also one of the founders of @black_in_ai, which has members around the world & has improved the entire field of AI. I attended #BlackinAI workshops in 2018 & 2019 and they were my favorite parts of NeurIPS

https://twitter.com/math_rachel/status/1071235139882156032?s=20

In particular, @Black_in_ai does a great job of covering the span of abstract technical advancements, practical applications, & concretely addressing societal impact, in a way that few other groups do.

Timnit (along with many others) put countless hours in to trying to help Africans get visas to NeurIPS, working 5 months in advance. This is such tangible, practical work to increase inclusion & try to address a terrible injustice (visa denials) in AI:

https://twitter.com/math_rachel/status/1072989099785576449?s=20

I love that @timnitGebru spoke about clique culture at CVPR 2018, about how hard it can be for outsiders to break into machine learning, how cliques harm diversity, & what we can do to be more welcoming:

https://twitter.com/math_rachel/status/1072989099785576449?s=20

Timnit describes her time at NeurIPS 2016, of seeing only 6 Black attendees out of 8,500: "I was literally panicking. This field was growing exponentially, hitting the mainstream; it’s affecting every part of society. It is an emergency, and we have to do something about it now."

Above quote is from "'We’re in a diversity crisis': cofounder of Black in AI on what’s poisoning algorithms in our lives" from MIT Tech Review, Feb 2018

technologyreview.com/2018/02/14/145…

Here's the link to Timnit's talk on countering clique culture:

https://twitter.com/math_rachel/status/1014309183087230976?s=20

Also, @timnitGebru is one of the founders of the Fairness, Accountability, & Transparency Conference (@FAccTConference), a major conference on ethics in machine learning:

https://twitter.com/geomblog/status/1334358000727810048?s=20

Timnit was also part of the team behind Model Cards for Model Reporting, to clarify intended use of an ML model, limitations, details of performance evaluation (including checking for bias), & more

academic paper: arxiv.org/abs/1810.03993

website: modelcards.withgoogle.com/about

She put in a huge amount of work advocating & organizing for ICLR (major machine learning conference) to be held in Ethiopia in 2020 (later cancelled due to covid), trying to counter the western-centric bias of ML confs

venturebeat.com/2018/11/19/maj…

Dr. Gebru's Tutorial on Fairness, Accountability, Transparency, & Ethics in Computer Vision from CVPR 2020

Slides & videos available here: sites.google.com/view/fatecv-tu…

https://twitter.com/roboticseabass/status/1334862114985553922?s=20

Closing the AI Accountability Gap: Defining an End-to-End Framework for Internal Algorithmic Auditing

arxiv.org/abs/2001.00973

https://twitter.com/DrLukeOR/status/1334798956958679040?s=20