Prior to joining Google in 2018, Gebru worked with MIT researcher Joy Buolamwini on a project called Nuances of gender who revealed face analysis IBM and Microsoft’s technology was very precise for white men but very imprecise for black women. It helped push us legislators and technologists to question and test the accuracy of facial recognition on different demographics, and Microsoft has helped, IBM, and Amazon announcing that they would suspend sales of technology this year. Gebru also co-founded an influential conference called Black in AI which attempts to increase the diversity of researchers contributing to this field.
Gebru’s departure was sparked when she collaborated with researchers inside and outside Google on an article discussing ethical issues raised by recent advances in AI language software.
Researchers have made considerable progress on issues like text generation and answer questions by creating a giant machine learning models trained on huge swathes of online text. Google says technology has made its eponymous search engine lucrative more powerful. But the researchers also showed that creating these more powerful models consumes large amounts of electricity due to the vast computing resources required, and documented how the models can reproduce biased language about gender and race found online.
Gebru says his draft paper discussed these issues and encouraged responsible use of technology, for example by documenting data used to create language models. She was troubled when the senior manager insisted that she and other Google authors either remove their names from the document or remove it altogether, especially when she couldn’t learn the process used to review the project. “I felt like we were being censored and I thought that had implications for all ethical research on AI,” she says.
Gebru says she failed to convince the senior manager to resolve the issues with the newspaper; she says the principal insisted that she withdraw her name. On Tuesday, Gebru responded by email to offer her a deal: if she received a full explanation of what had happened and the research team met with management to agree on a process for fair treatment of future research. , she would remove her name from the document. Otherwise, she would arrange to leave the company at a later date, leaving her free to publish the article without the company’s affiliation.
Gebru also emailed a larger list within Google’s AI research group, claiming managers’ attempts to improve diversity had been ineffective. She included a description of her dispute over the language document as an example of how Google officials can silence people from marginalized groups. Platform published a copy of the Thursday email.
On Wednesday, Gebru said, she learned from her direct reports that they had been told she had quit Google and her resignation had been accepted. She discovered that her corporate account was disabled.
An e-mail sent by a manager to Gebru’s personal address indicated that her resignation was to take effect immediately, as she had sent an e-mail reflecting “behavior incompatible with the expectations of a manager at Google”. Gebru took to Twitter and outrage quickly grew among online AI researchers.
Many Google critics, both inside and outside the company, noted that the company had suddenly undermined the diversity of its AI workforce and also had lost a leading advocate for the improvement of this diversity. Gebru suspects that his treatment was in part motivated by his outspokenness about diversity and Google’s treatment of people from marginalized groups. “We argued for representation, but there are hardly any blacks in Google Research and, from what I see, no leader at all,” she said.