undefined | Dark Hacker News

23 points by hughzhang 5 years ago

https://www.wired.com/story/behind-paper-led-google-research...

The odd thing is the paper itself is rather innocuous. Fairly standard stuff really. It seems clear to me that it was a pretext to force her out.

Many other Brain researchers (including on here) have indicated that it's quite common (perhaps ~40%) to submit for approval a day before conference submission, or even after submission. Given how many papers Google researchers submit, the intent was to give papers a once-over and make sure no IP was being disclosed. The fact that this paper went through an elaborate and secretive review process delivered by HR should tell us that this is, ultimately, not about the paper itself.

This has to do with a long-standing desire within some in the org to get rid of Gebru. Notably, this does not include Samy Bengio (research director, and her direct manager) who posted on Facebook his support and that he was kept out of the loop - I think there are tensions between the research teams and management also here.

However, by making it about the paper and the publication process, Google is certainly making it seem like they want to hide something (and perhaps they do..). Really goofed up on this. To the extent that Sundar Pichai had to write an email to all of Google about it today.

908B64B197 5 years ago |

I'm still not certain what she contributed to the field.

Most folks in the AI community learned of her existence when she rallied an angry mob at Yann LeCun, after he calmly tried to explain why certain results are biased [0].

Is this current thing with Google another personal branding/publicity stunt?

[0] https://syncedreview.com/2020/06/30/yann-lecun-quits-twitter...

iab 5 years ago |

Orthogonal to the who/what/why of this particular scenario, this is completely in keeping with the at-will employment philosophy. If the parties have irreconcilable differences, which is my impression from these articles, then negotiating a mutually-agreeable end-date seems like a no-brainer. Instead we have these immediate executive actions, which always end up benefiting (or at least not-harming) the employer.

nikivi 5 years ago |

Can someone give a tl;dr of what actually happened? I saw this incident mentioned many times from both the side of Timnit and Google and still have no idea why this happened.

benjohnson 5 years ago | |

My opinion:

A hard working researcher had the the flaw of not being able to produtivly work with some people. Some of those people were good and some of those people were bad.

At some point, some of those people decided that the cost of drama exceeded the value the work.

Much more drama ensued.

scared2 5 years ago | |

She had toxic relationship with her employer (google). She wrote a paper about Google's deep learning nlp google considered it a bad publicity. They asked her to include references for her claim. she gave precondition. if precondition is not met she would resign.

google took that line immediately and told her "your resignation is accepted".

that's the whole story.

iab 5 years ago | |

Differences of opinion on to-be published work, focusing (I believe) on bias in large-scale language models. Gebru was told to retract which was unacceptable, posed an ultimatum to Google (who accepted, immediately without negotiation) leading to her immediate termination.