AI researcher Timnit Gebru resigns from Google

TLDR : Yann Lecun tweeted that ML models are biased when the dataset they are trained on is biased. Bunch of people, including Timnit Gebru, replied he is wrong and the biases are every where and systemic.

Yann later tweeted again to explain what he meant in the context of deep learning models without hand engineered features, but people ignored him and still replied he was wrong.

It looks like they were accusing Yann of all the systemic biases of society... common.

Is there a paper showing that the Transformers based models are "biased" no matter the dataset ?

Nobody said it was disallowed. The grandparent said he called it out as a lazy rhetorical move. It's certainly true that oppression can be tiring (does that point need research?) It may equally be true that some people overclaim on the extent of their oppression, and the mental health harms they are suffering from it, and even make a profession of doing so.

IMHO she is saying it’s reasonable to believe that social and systemic issues have an impact on creating ML models (as the teams designing, building, and deploying said models are members of society) in ways in addition to the data sets.

I don’t know what the right answer is but plenty of institutions come up with ways they attempt to deal with systemic biases. For example, blind auditions for positions in orchestras (you only hear the person auditioning, you can’t see them). Used to not be a thing. Now it is.

Point being that when people are motivated they are able to come up with stuff.

I don't think she's saying she's marginalized because she's a well known researcher who people listen to, she's saying she's marginalized because one or more people, who have their identities concealed, asked her to retract a paper she already published via HR.

Granted, because the process she described is anonymous and opaque by design, it could have happened to other people at the company and only they, HR, and the people requesting/ordering the retraction would know about it. At the same time, if that's the case, then Google needs to stop that, because it looks really bad, especially when it involves a minority researcher who published a paper on how certain models Google uses may have some degree of discriminatory bias built into them.

Imagine this: You’ve sent a paper for feedback to 30+ researchers, you’re awaiting feedback from PR & Policy who you gave a heads up before you even wrote the work saying “we’re thinking of doing this”, working on a revision plan figuring out how to address different feedback from people, haven’t heard from PR & Policy besides them asking you for updates (in 2 months). A week before you go out on vacation, you see a meeting pop up at 4:30pm PST on your calendar (this popped up at around 2pm). No one would tell you what the meeting was about in advance. Then in that meeting your manager’s manager tells you “it has been decided” that you need to retract this paper by next week, Nov. 27, the week when almost everyone would be out (and a date which has nothing to do with the conference process). You are not worth having any conversations about this, since you are not someone whose humanity (let alone expertise recognized by journalists, governments, scientists, civic organizations such as the electronic frontiers foundation etc) is acknowledged or valued in this company.

Then, you ask for more information. What specific feedback exists? Who is it coming from? Why now? Why not before? Can you go back and forth with anyone? Can you understand what exactly is problematic and what can be changed?

And you are told after a while, that your manager can read you a privileged and confidential document and you’re not supposed to even know who contributed to this document, who wrote this feedback, what process was followed or anything. You write a detailed document discussing whatever pieces of feedback you can find, asking for questions and clarifications, and it is completely ignored. And you’re met with, once again, an order to retract the paper with no engagement whatsoever.

Jeff Dean's response about mitigations being made to minimize the issues she highlighted could be justified too, but again, because the process she described is opaque, no one external to the process has any idea about the merits of the arguments raised.

But the feedback itself wasn't shared with her. That part doesn't make sense unless management didn't want her to see it, and why would that be the case?