Senior leaders of Google's AI division held a meeting Tuesday to discuss the ousting of Timnit Gebru, a co-lead on Google's ethical AI team who announced last week that she had been fired by the company.
Last Wednesday, Gebru said she was terminated over a dispute regarding a research paper that she co-authored with several other Googlers. When Gebru sent an email to an employee resource group venting her frustration over the conflict, a manager used it as grounds to expedite her exit, she claimed.
The whole debacle has angered employees and members of the AI field, in which Gebru is a renowned name. More than 2,000 Google employees have signed a petition demanding leadership better explain Gebru's termination.
On Monday, Marian Croak, a VP of engineering and a senior Black Googler, invited employees in the company's AI division and those in the Black Googler Network to a virtual "community conversation" about Gebru's exit.
"News of Timnit Gebru's departure from Google has been very difficult for many of us. I wholeheartedly believe in the importance of both the research in this space and the need to have dialogue and provide clarity to the community at large," read the email from Croak, which was obtained by Business Insider.
Croak hosted the conversation, which was broadcast live to employees on Tuesday, and included Google's AI head Jeffrey Dean and Megan Kacholia, a research VP.
Tensions had been building inside Google for several days before Gebru's termination. Gebru had co-authored a research paper exploring bias in artificial intelligence and submitted it to a research conference. Following an internal review of the paper, Google asked that the paper be retracted or the names of the authors redacted.
Gebru offered some conditions to remove her name, one being to be informed of the names of everyone Dean and Kacholia had consulted during the review process. She said that she would "work on a last date" if Google could not meet these conditions, according to a tweet Gebru shared last week. Management refused, and informed Gebru that her resignation had been accepted. She said she wasn't given an opportunity to respond.
During Tuesday's conversation, Dean told employees that Gebru's ousting was not related to the content of the paper itself, which casts a critical light on some of the work Google and others are currently doing in the field of AI.
Instead, it was because of her condition of demanding the identities of the reviewers, he said, according to two employees who viewed the discussion and asked not to be identified because of fear of retaliation.
Dean said that there needs to be an opportunity to give feedback anonymously in research. "Sometimes those identities can't be revealed," he said, according to a portion of the transcript shared with Business Insider.
He also doubled down on his previous assertion that Timnit had offered her resignation, which Google had accepted. He referred to Timnit's conditions to know the identities of the paper's reviewers as a "binary choice."
It also was announced during the call that Croak will work on AI diversity research going forward, and will have a dotted report line to Dean, according to two employees who were on the call. Some employees balked at the decision, pointing out that Croak did not have a background in AI ethics, one of them said.
Employees were unable to provide comments or questions during the conversation, but were invited to a separate Q&A after the chat, which was held by an external diversity and inclusion consultant. The senior leaders did not attend that session.
During the Q&A, in which employees could "vote up" their most pressing questions, one Googler asked why leadership had not de-escalated the situation with Gebru, pointing out that Android creator Andy Rubin had been handed a $90 million exit package in his dismissal over sexual harassment claims. "How can we believe that our talent is of equal value at Google?" they asked.
In an anonymous Googler chat group, some employees aired their frustration after the meeting, according to messages viewed by Business Insider. "Kinda felt insulted by a lot of Jeff Dean's responses," read one. "We get it, anonymous feedback is an important part of the research review process. but why was her feedback given to her via HR? How come the other writers of the paper were not given feedback via HR if that was a critical part of the process? Why was her manager not included or even consulted in the decision of her *rolls eyes* 'resignation' Where are the assurances this won't happen again?
"I'm no longer incentivized to recruit other Blacks or underrepresented folks beyond stating the good pay," read another.
"I wish we didn't sell ourselves as a company who cared so deeply about DEI. This would sting so much less."
Some Google employees, including former colleagues of Gebru, were tweeting during the call to express their frustration at the comments made by leadership. "Employer is broadcasting that @timnitGebru put forward a binary approach that she would 'resign' if they didn't meet her demands,"tweeted Margaret Mitchell, another co-lead on the AI ethics team. "She did not."
In an email to employees last week, Dean defended the decision to terminate Gebru's employment. "We accept and respect her decision to resign from Google," he wrote.
But like Gebru, some employees dispute this claim, and on Monday published a letter claiming Gebru did not resign, but was fired. "Dr. Gebru's dismissal has been framed as a resignation, but in Dr. Gebru's own words, she did not resign," the letter read.
A Google spokesperson did not respond to Business Insider's request for comment.