6 ms·

She wasn't asked to change it, and she didn't refuse. She was summoned to a meeting where she was informed that the paper must be retracted by her, due to subs

She wasn't asked to change it, and she didn't refuse. She was summoned to a meeting where she was informed that the paper must be retracted by her, due to substantive feedback that was provided anonymously, that would not be provided to her. She ask to be given the feedback and a chance to address it so that the publication process could continue. After some discussion, she was told her manager would be allowed to read her the feedback, but she wouldn't be provided a copy; regardless, the paper must be retracted. She was not given an opportunity to edit the paper based on the feedback. This was the cause of her email to the list. Google hasn't disputed any of these details.

ALittleLight 6y ago
Jeff Dean's document makes it sound like they aren't opposed to Gebru's paper in principle, but are opposed to it without what they consider missing context. "A cross functional team then reviewed the paper as part of our regular process and the authors were informed that it didn’t meet our bar for publication and were given feedback about why. It ignored too much relevant research — for example, it talked about the environmental impact of large models, but disregarded subsequent research showing much greater efficiencies. Similarly, it raised concerns about bias in language models, but didn’t take into account recent research to mitigate these issues. We acknowledge that the authors were extremely disappointed with the decision that Megan and I ultimately made, especially as they’d already submitted the paper. " And: "But the paper itself had some important gaps that prevented us from being comfortable putting Google affiliation on it. For example, it didn’t include important findings on how models can be made more efficient and actually reduce overall environmental impact, and it didn’t take into account some recent work at Google and elsewhere on mitigating bias in language models. Highlighting risks without pointing out methods for researchers and developers to understand and mitigate those risks misses the mark on helping with these problems. As always, feedback on paper drafts generally makes them stronger when they ultimately appear." I'd be surprised, having read this, if Gebru and coauthors made the requested adjustments, the paper wouldn't have been approved by Google. That's my intuition at least, maybe I'm off. https://docs.google.com/document/d/1f2kYWDXwhzYnq8ebVtuk9CqQqz7ScqxhSIxeYGrWjK0/edit?usp=drivesdk https://docs.google.com/document/d/1f2kYWDXwhzYnq8ebVtuk9CqQ...
- fatbird 6y ago
  Gebru's account is that she was first refused the feedback, and then offered only a recitation of the feedback, not the written account. She offered to rewrite to account for the feedback, and was denied. Dean doesn't dispute this.
  - ALittleLight 6y ago
    Has she given her account in a linear way? (i.e. not on Twitter)
- snowwrestler 6y ago
  Dean’s explanation makes sense if it is based on a standard that is consistently applied to all research papers coming out of Google. But there are public statements from other Googlers implying that it’s not, as noted in this article: https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru/ https://www.technologyreview.com/2020/12/04/1013294/google-a...
- axaxs 6y ago
  This is in line with my entire problem with the whole Twitter fiasco: screaming about problems but not offering any solutions. I don't know the lady well, but given her background I assume she's intelligent. So, this isn't to degrade her but that trait in general - in what job can your role be to just tell people that they're wrong? How long would a roofer last that just stood on the ground and yelled up they were doing it wrong...even if the roofers really were? Or a programmer who denies your PR because it's wrong? Communication is an invaluable life skill. Handling disagreements with tact and real suggestions in hand will always be better received.
  - loopz 6y ago
    You ALWAYS must provide a constructive suggestion. Only criticizing shuts people down. However, rotten leadership can't handle anything less than full loyalty, and this is rampant in companies, and now also politics.

ALittleLight 6y ago

Jeff Dean's document makes it sound like they aren't opposed to Gebru's paper in principle, but are opposed to it without what they consider missing context. "A cross functional team then reviewed the paper as part of our regular process and the authors were informed that it didn’t meet our bar for publication and were given feedback about why. It ignored too much relevant research — for example, it talked about the environmental impact of large models, but disregarded subsequent research showing much greater efficiencies. Similarly, it raised concerns about bias in language models, but didn’t take into account recent research to mitigate these issues. We acknowledge that the authors were extremely disappointed with the decision that Megan and I ultimately made, especially as they’d already submitted the paper. " And: "But the paper itself had some important gaps that prevented us from being comfortable putting Google affiliation on it. For example, it didn’t include important findings on how models can be made more efficient and actually reduce overall environmental impact, and it didn’t take into account some recent work at Google and elsewhere on mitigating bias in language models. Highlighting risks without pointing out methods for researchers and developers to understand and mitigate those risks misses the mark on helping with these problems. As always, feedback on paper drafts generally makes them stronger when they ultimately appear." I'd be surprised, having read this, if Gebru and coauthors made the requested adjustments, the paper wouldn't have been approved by Google. That's my intuition at least, maybe I'm off. https://docs.google.com/document/d/1f2kYWDXwhzYnq8ebVtuk9CqQqz7ScqxhSIxeYGrWjK0/edit?usp=drivesdk https://docs.google.com/document/d/1f2kYWDXwhzYnq8ebVtuk9CqQ...

fatbird 6y ago
Gebru's account is that she was first refused the feedback, and then offered only a recitation of the feedback, not the written account. She offered to rewrite to account for the feedback, and was denied. Dean doesn't dispute this.
- ALittleLight 6y ago
  Has she given her account in a linear way? (i.e. not on Twitter)
snowwrestler 6y ago
Dean’s explanation makes sense if it is based on a standard that is consistently applied to all research papers coming out of Google. But there are public statements from other Googlers implying that it’s not, as noted in this article: https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru/ https://www.technologyreview.com/2020/12/04/1013294/google-a...
axaxs 6y ago
This is in line with my entire problem with the whole Twitter fiasco: screaming about problems but not offering any solutions. I don't know the lady well, but given her background I assume she's intelligent. So, this isn't to degrade her but that trait in general - in what job can your role be to just tell people that they're wrong? How long would a roofer last that just stood on the ground and yelled up they were doing it wrong...even if the roofers really were? Or a programmer who denies your PR because it's wrong? Communication is an invaluable life skill. Handling disagreements with tact and real suggestions in hand will always be better received.
- loopz 6y ago
  You ALWAYS must provide a constructive suggestion. Only criticizing shuts people down. However, rotten leadership can't handle anything less than full loyalty, and this is rampant in companies, and now also politics.