Google AI scientist claims she was fired for highlighting bias in new paper

4 Dec 2020

Dr Timnit Gebru speaking at TechCrunch Disrupt 2018. Image: Kimberly White/Getty Images for TechCrunch

Dr Timnit Gebru, a black woman and prominent researcher into ethics in AI, said she was fired from Google for criticising the company’s diversity efforts.

Writing in an email to staff, a Google manager denied that the company fired its now former AI scientist and co-lead of its Ethical AI team, Dr Timnit Gebru, saying that she had offered her resignation.

Gebru claimed on Twitter on 3 December that she was fired from the company for criticising its diversity efforts in an email to colleagues. She also criticised the company over its approach to hiring minorities and not doing enough to stamp out biases in AI systems.

In the email – seen by The New York Times – she said she was frustrated with the lack of gender diversity in Google’s AI division and questioned whether the company was more critical of her efforts than those of different backgrounds.

“Your life starts getting worse when you start advocating for underrepresented people. You start making the other leaders upset,” Gebru wrote to colleagues. “There is no way more documents or more conversations will achieve anything.”

Gebru, along with fellow AI researcher Joy Buolamwini, was behind the landmark Gender Shades paper published in 2018 that – among other issues – showed some facial recognition systems were misclassifying black women as men.

Gebru was set to present a new paper co-written with both Google and non-Google staff at a computer science conference in March next year which, according to her, was not well received at Google. The paper, seen by The New York Times, included claims that a new type of language technology was created to tap into its search engine and scan hundreds of thousands of websites.

Respected ‘her decision to resign’

As many of these sites would include biased and hateful language, the paper said, the technology may end up generating significantly biased and hateful language. According to The New York Times, Google demanded that she either retract her name from the paper or pull it entirely, which Gebru refused.

After stating that she would resign after a certain amount of time if Google did not state why it wanted her to retract the paper, the company said it would not meet her demands and accepted her offer of resignation.

Gebru said that she was then immediately closed off from accessing any Google systems without warning or any conversations to confirm the resignation.

The head of Google’s AI unit, Jeff Dean, wrote an email to staff saying that the company respected “her decision to resign”. Dean also claimed that Gebru’s paper “ignored too much relevant research” that he claimed has shown improvements in AI made in recent years.

‘The amount of disrespect is incredible’

Responding to news last week that Google rejected her paper, Gebru posted on Twitter: “Nothing like a bunch of privileged white men trying to squash research by marginalised communities for marginalised communities by ordering them to stop with zero conversation. The amount of disrespect is incredible.”

Aside from Dean’s email being made public, Google has refused to officially comment on the matter.

Following the news, almost 1,000 Google staff and supports in academics and industry signed a letter of support for Gebru saying that her being cut off from Google has “overthrown a working understanding of what kind of research Google will permit” and is a “moment of reckoning” for the company.

“As we know, Dr Gebru is one of the few people exerting pressure from the inside against the unethical and undemocratic incursion of powerful and biased technologies into our daily lives,” wrote the Google Walkout for Real Change group. “This is a public service, and its importance cannot be overstated. Google’s retaliation against Dr Gebru, and its move to silence this work, concerns us all.”

Colm Gorey was a senior journalist with Silicon Republic