We had 128 citations [on that paper], and we sent our paper to many of these people that we cited. We were so thorough. I said, okay, I want to bucket the people that we’re going to ask feedback from in four buckets. One is the people who have developed large language models themselves, just to get their perspective. One is people who work in the area of understanding and mitigating the bias in these models. One is people who might disagree with our view. One is people who use these large language models for various things. And we have a whole document with all of this feedback that we were supposed to go through to address, and which I want to do still before we release this work.
But the way they [Google leadership] were talking to us, it wasn’t like they were talking to world-renowned experts and linguists. Like Emily Bender [a professor at the University of Washington and a coauthor of the paper] is not some random person who just put her name on a random paper out there. I felt like the whole thing was so disrespectful.
Prior to this particular paper, were there earlier instances in which you ever felt that Google was restricting or limiting your team’s ability to conduct research?
I felt like there were prior instances where they watered down the research results a lot. People had conversations with PR and policy or whatever, and they would take issue with certain wording or take issue with certain specifics. That’s what I thought they might try to do with this paper, too. So I wrote a document, and I kept asking them, “What exactly is your feedback? Is your feedback to add a section? To remove? What does this mean? What are you asking us?”
This was literally my email on Friday after Thanksgiving [November 27, five days before Gebru’s dismissal] because on Thanksgiving day I had spent my day writing this document instead of having a good time with my family. The next day, on Friday, which is when I was supposed to retract this paper, I wrote: “Okay, I have written this six-page document addressing at a high level and low level whatever feedback I can gather. And I hope that there is at the very least an openness for further conversation rather than just further orders.” I wrote that email. Like that. How does Megan [Kacholia, the VP of engineering at Google Research] respond to this email? Monday, she responds to it and says, “Can you please confirm that you have either retracted the paper or taken the names of the authors out of this paper. Thank you. And can you please confirm after you’ve done this. Send me an email and confirm.” As if I have no agency.
That’s not what they do to people who’ve engaged in gross misconduct. They hand them $80 million, and they give them a nice little exit. They don’t do what they did to me.
Then in that document, I wrote that this has been extremely disrespectful to the Ethical AI team, and there needs to be a conversation, not just with Jeff and our team, and Megan and our team, but the whole of Research about respect for researchers and how to have these kinds of discussions. Nope. No engagement with that whatsoever.
I cried, by the way. When I had that first meeting, which was Thursday before Thanksgiving, a day before I was going to go on vacation—when Megan told us that you have to retract this paper, I started crying. I was so upset because I said, I’m so tired of constant fighting here. I thought that if I just ignored all of this DEI [diversity, equity, and inclusion] hypocrisy and other stuff, and I just focused on my work, then at least I could get my work done. And now you’re coming for my work. So I literally started crying.
What do you think it was about this particular paper that touched off these events? Do you think that it was actually about this paper, or were there other factors at play?
I don’t know. Samy was horrified by the whole thing. He was like, this is not how we treat researchers. I mean, this is not how you treat people in general. People are talking about how, if this happens to somebody this accomplished—it makes me imagine what they do to people, especially people in vulnerable groups.