Google fires AI researcher Timnit Gebru over critical email

Google officials did not respond to requests for comment. Gebru also did not respond to requests, but in an interview Thursday with Bloomberg said Google’s actions represented “the most fundamental silencing.” Gebru said Wednesday that she had been fired by Jeff Dean, the head of Google’s AI division, for an email she had sent to Brain Women and Allies, a mailing list for researchers on the company’s AI team known as Google Brain.

Her dismissal threatens to reignite anger over Google’s treatment of its workforce, particularly its employees of color. The company, which long evangelized its open culture, has cracked down on employee dissent over the past two years, particularly against marginalized and minority workers. Google has previously fired employees who advocated for increased diversity or critiqued the company’s ethics.

Inioluwa Deborah Raji, an AI researcher on a fellowship with the Mozilla Foundation who has worked with her, said Gebru had helped lend Google an air of legitimacy and credibility as it worked to portray itself as committed to inclusion and ethical research.

Despite Google’s reputation as a leader in AI ethics within the tech industry, its efforts have been marred by controversy and internal criticism. It disbanded its AI Ethics Council, an external advisory board, weeks after launching in March 2019 when employees protested the inclusion of Heritage Foundation president Kay Cole James, who had no AI experience and had recently shared anti-immigrant and anti-transgender views.

Gebru occupied a rare role in a Silicon Valley culture long criticized for its racial homogeneity. As an Ethiopian American woman in a senior role at the tech giant, she became well known for work that critically examined the technology’s societal biases and repercussions.

She co-founded Black in AI, an advocacy group that has held workshops at major AI conferences and pushed for more Black roles in AI development and research. She has also regularly criticized tech companies, including Google, for failing to hire more workers of color and treating them differently once they’re on board.

Two days before announcing her firing, Gebru had solicited advice regarding whistleblower-like protections for AI-ethics researchers, tweeting, “With the amount of censorship & intimidation that goes on towards people in specific groups, how does anyone trust any real research in this area can take place?”

Tensions between Gebru and the company also stemmed from research by Gebru’s team that was critical of AI systems, known as large language models, said one machine-learning researcher who had reviewed the study and requested anonymity because they were not authorized to discuss the unpublished work. The company may one day seek to capitalize on such systems in consumer-facing products that could generate convincing passages of text that are difficult to distinguish from human writing, the researcher said.

Google already uses language-processing AI systems to translate between languages, look for hate speech and understand people’s writing and speech online. But other firms have increasingly sought to push the technology forward with systems, like the AI lab OpenAI’s GPT-3 system, which has been used to generate seemingly human-created news reports, poems and computer code.

Gebru told Bloomberg that the company had asked for her to remove Google employees’ names from, or retract outright, the research paper on language models, which was to be submitted for an AI conference next year.

In her email to the mailing list, first published by Casey Newton in his newsletter Platformer, Gebru described her treatment at Google as dehumanizing. “Have you ever heard of someone getting ‘feedback’ on a paper through a privileged and confidential document to HR? Does that sound like a standard procedure to you or does it just happen to people like me who are constantly dehumanized?”

Gebru recounted her most recent experience in the email as an example of why she had given up on advocating for diversity inside Google. “[S]top writing your documents because it doesn’t make a difference,” she wrote. “[Y]our life gets worse when you start advocating for underrepresented people, you start making the other leaders upset when they don’t want to give you good ratings during calibration.”

Brain Women and Allies is what’s known as an employee resource group, popular in the tech industry and elsewhere in corporate America, commonly organized by co-workers who share the same background and interest. In Silicon Valley, which is still largely dominated by White and Asian men, these voluntary groups have been a vital resource for advice, candor, and solidarity for marginalized employees, and one of the few internal arenas where workers may feel able to express the types of critiques Gebru shared in her email.

Gebru said her managers told her that the email had reflected “behavior that is inconsistent with the expectations of a Google manager,” and that they were accepting her resignation — even though, she said, she had not directly offered one. Gebru said she is currently on vacation and that the company had immediately terminated her access to company email accounts.

Gebru’s dismissal triggered an online outcry from fellow AI researchers such as Joy Buoalamwini, who tweeted that Gebru had asked “hard questions not to advance yourself but to uplift the communities we owe our foundations,” and that she had faced “institutions [that] try to cut your knees into submission.”

Alex Hanna, a researcher on Gebru’s team, tweeted that Dean was “now emailing the whole of the research organization, spreading misinformation and misconstruals about the conditions” of Gebru’s firing. She added, “Google researchers: don’t buy it.”

In a companywide email Thursday, first published in Platformer, Dean urged employees to continue working on Google’s diversity, equity and inclusion efforts. “Given Timnit’s role as a respected researcher and a manager in our Ethical AI team, I feel badly that Timnit has gotten to a place where she feels this way about the work we’re doing,” he wrote. “I also feel badly that hundreds of you received an email just this week from Timnit telling you to stop work on critical DEI programs. Please don’t. I understand the frustration about the pace of progress, but we have important work ahead and we need to keep at it.”

Ifeoma Ozoma, a former public policy executive for Pinterest who quit the company in May after she faced retaliation for raising complaints about pay inequity and race bias against Black employees, pointed out that Google defended former employee Miles Taylor, the ex-Department of Homeland Security official who worked on the Trump administration’s family separation policy. “Google went out of its way to defend his human rights violations, but fired her over not retracting her name from a paper and then emailing allies about it,” Ozoma said.

The company has sought to address criticism by releasing statements about its work to increase Black representation in its executive ranks. But Gebru’s dismissal, Raji said, threatens to undermine all of that, suggesting the company may be more committed to moneymaking commercial endeavors than harsher critiques of the technology’s real-world use.

“People would think, ‘Timnit’s there, so it’s evidence that there’s an openness I’d have to work with people there on these issues,’” Raji said. “Firing her in such a disrespectful way reveals that perhaps Google’s commitment to some of these issues was not as legitimate as previously believed.”

In May, less than a week after George Floyd, an unarmed Black man, was killed while in police custody, Gebru publicly stood up to Dean. The Google executive had tweeted his support for law enforcement and the idea that policy brutality was not universal. “You might think something is perfectly fine but you’re not the one dealing with this. This adds to the violence against us,” Gebru replied on Twitter, adding, “Most people would not even say anything to you.”

Source Article