Timnit Gebru

Some people arrive in a field and simply confirm what is already known. Others show up, look at the data with fresh eyes and say: "wait, this is wrong and someone needs to say it." Timnit Gebru is one of the latter. Born in Addis Ababa, Ethiopia in 1983, she came to the United States as a political refugee in 1999 and became one of the most important voices in modern AI, not because she built the largest model, but because she had the courage to ask who these systems are actually built for.

In 2018, together with Joy Buolamwini at the MIT Media Lab, she published the "Gender Shades" study: the first peer-reviewed empirical evidence that commercial facial recognition systems already deployed in security and business failed far more often on dark-skinned women than on light-skinned men. In some cases, the error rate reached 34.7% for Black women, compared to less than 1% for white men. This was not a hypothesis; those were numbers that companies could not ignore, though they tried. That single paper shook the industry and led Amazon, Microsoft, and IBM to review or pause the sale of their facial recognition systems to law enforcement.

Timnit earned her PhD at the Stanford AI Lab, then worked in the FATE (Fairness, Transparency, Accountability and Ethics) group at Microsoft Research in New York before joining Google, where she rose to co-technical lead of the Ethical AI team. Her departure from Google in December 2020 was public and painful: the company asked her to retract a paper on the risks of giant language models, and when she refused, the situation escalated to her firing. Instead of going quiet, she spoke out, and the world listened. In 2021 she founded the DAIR Institute (Distributed AI Research Institute), an independent lab dedicated to documenting the impact of AI on marginalized communities, with a particular focus on Africa and Africans in the diaspora. She is also a co-founder of Black in AI, the group that opened doors for Black researchers in an industry that had long overlooked them.

The lesson I take from Timnit is that technical truth and social justice are not separate paths: they are the same path. Saying "this system fails more often with Black people" is not politics, it is science. And it takes people willing to do that science even when it is uncomfortable. Every time you use an AI tool today, there is a greater chance it is a little more fair because of her work. That is real legacy.