24/05/2026
Addis Ababa, Ethiopia. Around 1983.
Timnit Gebru was born into a household that understood what education could do. Her father held a PhD in electrical engineering. Her mother was an economist. Both of her parents were from Eritrea. When Timnit was five years old, her father died.
Her mother raised her and her siblings alone, carefully and with complete conviction about one thing. Education was the one possession nobody could take from you. Books were stacked on every surface. Learning was the household religion.
Then in 1998, when Timnit was fifteen years old, the Eritrean-Ethiopian War erupted and changed everything. Because of her family's Eritrean roots, they became targets. Some of her family members were deported to Eritrea and forced into military service. Timnit fled Ethiopia, leaving behind everything she had known, and eventually arrived in the United States to apply for political asylum.
She was denied initially.
She kept fighting.
She was eventually granted asylum and settled in Massachusetts, where she enrolled in high school. She arrived barely speaking fluent English. School administrators, seeing a refugee teenager whose first language was not English, attempted to place her in lower-level classes.
She fought her way into the advanced ones. Then she became one of the strongest students in the building.
An experience during those years in Massachusetts would later shape the entire direction of her career. A friend of hers was assaulted. When she contacted the police, the police arrested her friend instead of helping them. Timnit watched a system that was supposed to protect people harm an innocent person instead, and she could not stop thinking about why that happened and how it could be prevented.
In 2001, she enrolled at Stanford University. She earned a bachelor's degree in electrical engineering. Then a master's. Then a PhD from Stanford's Artificial Intelligence Laboratory, completing her doctorate in 2017, with her research focused on computer vision, machine learning, and a methodology she developed for using Google Street View images to predict demographic data and voting patterns in American neighborhoods.
She did a postdoctoral fellowship at Microsoft Research in New York, working in the FATE group, studying algorithmic bias and the ethical implications of data-driven systems. She co-founded Black in AI, an advocacy organization pushing for more Black researchers and practitioners in artificial intelligence. She was building something, piece by piece, with the particular focus of someone who understands that systems can cause harm at scale and that understanding exactly how requires careful, methodical work.
Then in 2018 came the paper that changed her field.
Working with MIT researcher Joy Buolamwini, Timnit Gebru published Gender Shades, a study that tested facial recognition systems being sold by major technology companies to police departments and governments around the world.
The results were not an opinion. They were measurements.
For light-skinned men, the error rate was less than one percent.
For dark-skinned women, it reached 34.7 percent.
The same technology being used to identify criminal suspects, verify identities, and surveil communities was failing dramatically for the people with the least power to challenge it when it got things wrong. The study won an AI for Good award in 2019 and became a landmark document in the field of AI ethics, forcing the technology industry to confront algorithmic bias not as a future concern but as a documented, measurable harm happening right now, at scale, in systems already deployed.
That same year, Google hired her as a research scientist and she became technical co-lead of their Ethical AI team, one of the most respected and diverse AI research groups at any major technology company.
She was doing exactly what they hired her to do.
In 2020, Gebru and several colleagues wrote a paper titled "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?" It raised three serious concerns. Large language models learn from billions of words scraped from the internet. That text carries racism, sexism, and prejudice baked into everyday language across decades. The AI absorbs those patterns and repeats them at massive scale, into hiring systems and chatbots and loan decisions, before anyone fully understands what is happening. Training a single large model can produce carbon emissions equivalent to the lifetime output of five cars. And the datasets are so enormous that no team can fully audit them. No one can guarantee who they quietly harm.
The paper did not demand that AI stop. It asked the industry to slow down. To think carefully. To ask who gets hurt when we rush.
Google did not want to slow down.
They asked her to retract the paper. Or remove her name. She asked for transparency, for specific feedback, for the honest conversation the company publicly claimed to value. She requested that Google disclose more information about its approach to reviewing diversity shortcomings.
On December 2, 2020, Google ended her employment. The company initially framed it as a resignation she had never made.
More than 2,700 Google employees signed a public protest letter. Over 4,300 researchers and academics around the world joined them. Nine members of the United States Congress sent a letter directly to Google's CEO demanding an explanation.
Timnit Gebru did not wait for one.
Exactly one year after her firing, on December 2, 2021, she launched DAIR, the Distributed Artificial Intelligence Research Institute, funded by the MacArthur Foundation, the Ford Foundation, the Kapor Center, the Open Society Foundation, and the Rockefeller Foundation to the tune of $3.7 million. Researchers across Africa, Europe, North America, and Australia, accountable to no shareholder and no quarterly earnings report, doing precisely the work Google had hired her to do, freely and without institutional interference.
One of DAIR's first projects used AI to analyze satellite images of South African towns to study the long-term effects of apartheid on communities. The work was exactly what she had always said AI could be used for, understanding and addressing the harm that systems cause rather than accelerating the harm they are capable of.
Fortune named her one of the world's 50 Greatest Leaders. Nature listed her among the scientists who shaped the year. Time named her one of the 100 Most Influential People in the World.
Her firing also helped ignite the formation of the first union by tech workers at Google, a labor organizing effort that drew a direct line from her case to the broader question of whether the people building these systems have any power to shape how they are used.
Think about that arc.
Her father died when she was five. She fled a war at fifteen. School administrators tried to place her in remedial classes. She earned a PhD from Stanford. She proved with measurements that facial recognition technology failed dark-skinned women at thirty-four times the rate it failed light-skinned men. She joined the most powerful technology company on earth to make AI safer. She told the truth about what she found.
They fired her.
She built something they could not control.
Right now, AI is making decisions that shape real lives. Who gets a job interview. Who gets approved for a loan. Who gets flagged by a security system. Those systems are trained on historical data full of historical bias. They amplify it. They automate it. They scale it to millions of decisions before anyone fully understands what is happening.
Someone has to slow down long enough to ask who is being hurt.
Timnit Gebru keeps asking.
Whether Silicon Valley wants to hear it or not.
She came to America with nothing but the belief that education could carry a person through almost anything.
Turns out she was right.
~The History Today