Timnit Gebru

13 results

Screenshot of an image from the linked Fast Company article in the description. The image has 3 circles, the first circle has a photo of Timnit, the 2nd is a green circle and the 3rd is a yellow circle that looks like it has circuitry. The background is black, with what looks like a fence.

Many models for many people

Counteracting the "one giant model for everything" approach by BigTech, by empowering small, community-rooted organizations around the world building task specific models serving their needs. Read this interview and our paper.

An image by Jamillah Knowles & Reset.Tech Australia / Better Images of AI / People with phones / CC-BY 4.0 It shows a bunch of people walking, standing in a wheel chair, all looking at their phones. The image is brown in color.

Social media harms

Creating natural language processing tools, and using mixed methods (quantitative and qualitative) approaches to analyze the impacts of social media platforms on neglected countries and languages. Watch our short documentary on the topic.

Top-view image of townships on the Left. DAIR Fellow Raesetja Sefala pictured on the right, seated and looking left..

Africa

Research

Tools

Impacts of spatial apartheid

Analyzing the impacts of South African apartheid using computer vision techniques and satellite imagery. Read our NeurIPS paper and this MIT Tech Review article, and check out our dataset and visualizations.

Painting of two knowledge workers, running inside a wheel embedded within a computer mouse, struggling to keep pace as a powerful hand (representing employers or capitalistic forces).The image reveals that despite AI and digital adoption, rising productivity expectations often trap workers into producing more with less. Rather than easing labour, AI technologies risks raising the baseline demands, accelerating the pace of work under the illusion of progress. This image was selected as a winner in the Digital Dialogues Art Competition, which was run in partnership with the ESRC Centre for Digital Futures at Work Research Centre (Digit) and supported by the UKRI ESRC.o

AI-fueled inequities

AI is fueled by the artists, writers and “zombie trainers” (see below) supplying training and evaluation data without consent or compensation. Vulnerable populations like refugees and gig workers are hired as data workers under exploitative working conditions (see Data Workers Inquiry). These same vulnerable populations are also the first to experience the harms of AI systems, whether it is through automated weapons or surveillance systems (see Surveillance Watch). This page has more of our research on the inequities fueled by AI systems.

Image: Leo Lau & Digit / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/

TESCREAL

The TESCREAL Bundle

Big tech companies are throwing money at artificial general intelligence (AGI), with CEOs telling us that it will come, bringing with it utopia, any day now (for the measly price of SEVEN TRILLION DOLLARS according to Sam Altman). How did AGI come to be a dominant goal in AI? What motivates the drive to attempt to build it? This page is dedicated to answering this question, starting with our paper, "The TESCREAL Bundle: Eugenics and the Promise of Utopia through Artificial General Intelligence," which you can read on First Monday. Image is Clarote & AI4Media / Better Images of AI / Power/Profit / CC-BY 4.0

A colorful slide of what looks like post its. With "Who is who in data labeling?" first and then a bunch of sketches of people. Most of the text in the post its is not legible. But there is a title, "Stakeholders Map"

Research Frameworks

Documentation & accountability

Guidelines for ethical data annotation and release practices without exploiting workers and stealing data.

An image of a butterfly perched on a branch, framed in a square, surrounded by seemingly energised earthy elements which are seen in a pixellated form.

Research Frameworks

Need-based design of AI systems

Creating guidelines for designing AI systems that center the needs of specific communities.

Screenshot from the Huniki Federation website features abstract designs in bright primary colors like blue, green, and red, on a white background. The text reads, "Outperform the mega-models. Put language tech in the hands of people who actually speak it." The logos for participating organization line up below this: Khaya AI, Lesan AI, Algorine, Ghana NLP, Lelapa, and DAIR.

Language Technology

Africa

Tools

Huniki Federation: Many Models for Many People

Counteracting the "one giant model for everything" approach by BigTech by empowering small, community-rooted organizations around the world building task-specific models serving their needs. Check out the Huniki Federation website for more.

Events

Celebrating Our Second Anniversary

With your support, the last two years have helped us strengthen this ecosystem and we’re already starting to see a shift in the voices that shape AI research practices.

A yellow window with three images. On the right is an under water photo of a single bright blue and yellow tropical fish. The two on the left are simplifications created based on a decsion tree algorithm. The work illustrates a popular type of machine learning model: the decision tree. Decision trees work by splitting the population into ever smaller segments. I try to give people an intuitive understanding of the algorithm. I also want to show that models are simplifications of reality, but can still be useful, or in this case visually pleasing.

News

Statement from the listed authors of Stochastic Parrots on the “AI pause” letter

The harms from so-called AI are real and present and follow from the acts of people and corporations deploying automated systems. Regulatory efforts should focus on transparency, accountability and preventing exploitative labor practices.

Illustration of four people clustered together, looking at a laptop screen. They have a variety of skin tones, but all of them have the appearance of age -- gray streaks in hair, white hair, or wrinkled faces. Image pop-outs from the computer show the faces of two children who they appear to be talking to. The wall behind the elders is purple, and a plant sits on a stool to the left.

Possible Futures

Blog

An Internet for Our Elders

An internet that benefits our future is one that enables everyone to contribute to the richness of our global conversation.

Events

Human Rights

Video

#NoTechForApartheid

Thank you to everyone who joined us for our #NoTechForApartheid panels, co-hosted with Haymarket Books. If you missed the event, you can find the recording on Peertube. We learned a lot, and we are certain that you will too.

A complicated sketch which looks like it was summarizing Mila's talk. Most of the text is not legible as a thumnail. At the top it says Welcome Everyone. Then "Dataset Documentation" with sketches of a number of people. What looks like a calendar, and then a "1. Stake Holder Map" Another image which looks like a child fallen from a bike. Then Day 3, and a sketch of someone in front of a white board. Then "Challenge #2" and illegible text. Then Challenge #3 and Future: What? How? Result