4 ms·

as they say in eastern europe "Don't spit in the well you drink from" this reminds of that Google ethics researcher who was fired for sabotage and then built a

as they say in eastern europe "Don't spit in the well you drink from" this reminds of that Google ethics researcher who was fired for sabotage and then built a big following on Twitter as a professional "victim of racism"

whymauri 5y ago
Timnit had a large following before she got fired. Especially for us who work in the field. She is not a 'professional victim of racism.' She is a professional who was fired for raising ethical concerns around larger language models.
- dodobirdlord 5y ago
  > She is a professional who was fired for raising ethical concerns around larger language models. This is a clear mischaracterization of the reported events. Gebru, in her own statements on the subject, stated that she was fired for demanding to know the identities of peers who submitted critical feedback on her work through confidential HR channels. Leaving aside the question of whether those criticisms had merit, her being fired was a foregone conclusion of making such an ultimatum. No confidential peer feedback system can function if there is a possibility of reviewers being de-anonymized at the demand of sufficiently influential coworkers. On a more opinionated note - making a demand like that displays critical lack of judgement in the important responsibility that all employees in supervisory roles have in ensuring a non-hostile working environment. Even if her intention was not to retaliate against her peers after determining their identities, it’s essential that people in positions of power do not take actions that even suggest that they are seeking to retaliate against less powerful peers. The importance of this principle is extreme to a healthy collaborative work environment. The fact that someone as intelligent as Gebru was surprised to be fired for making an ultimatum like this is itself surprising.
danielheath 5y ago
Sabotage is a pretty wild way to say “refusing to retract an academic paper that points out obvious flaws in ML models”.
- andyxor 5y ago
  there was much more to that, she demanded a list of reviewers who rejected her paper in a sort of ultimatum, and blasted group emails calling for sabotage I don't know if you saw Jeff Dean's response, but here it is https://docs.google.com/document/d/1f2kYWDXwhzYnq8ebVtuk9CqQqz7ScqxhSIxeYGrWjK0/preview?pru=AAABdlOOKBs*gTzLnuI53B2IS2BISVcgAQ https://docs.google.com/document/d/1f2kYWDXwhzYnq8ebVtuk9CqQ... There was also a thread on reddit with her colleagues reactions https://www.reddit.com/r/MachineLearning/comments/k77sxz/d_timnit_gebru_and_google_megathread/?sort=top https://www.reddit.com/r/MachineLearning/comments/k77sxz/d_t...
- dekhn 5y ago
  you left out "publicly calling out your SVP on twitter for not being racially sensitive enough" which she did before the paper blowup. Pretty sure that was the prime impetus for her being fired. BTW, I worked with TPUs on the hardware side, and the entire section about hardware power usage is just completely and totally wrong (I would have asked for the paper to be retracted if it was published, the errors are so large). As for the obvious flaws part, yeah, some of the flaws are obvious (and known to all practitioners of the art already), some are irrelevant but made to sound like they are existential crises, and some of the "flaws" are just per personal opionion about what she doesn't like. Hopefully she can find some academic position from which to propound her opinions. However, I do think in her behavior, she excluced herself from future employment.
  - cdprojektdead 5y ago
    Multiple other, even white, employees called out this same SVP. Your usage of the phrase "Not being racially sensitive enough" strikes me as a bizarre euphemism for "racism", which it would appear you don't see as an issue. "are just per personal opinion about what she doesn't like." The field of AI Ethics is an amalgam of computer science and moral philosophy. What do you think value judgements are? What are those "personal opinions" and things she "doesn't like" and why? It seems you're using a lot of euphemisms to hide some very unpleasant truths about your world view.
    - 5y ago

whymauri 5y ago

Timnit had a large following before she got fired. Especially for us who work in the field. She is not a 'professional victim of racism.' She is a professional who was fired for raising ethical concerns around larger language models.

dodobirdlord 5y ago
> She is a professional who was fired for raising ethical concerns around larger language models. This is a clear mischaracterization of the reported events. Gebru, in her own statements on the subject, stated that she was fired for demanding to know the identities of peers who submitted critical feedback on her work through confidential HR channels. Leaving aside the question of whether those criticisms had merit, her being fired was a foregone conclusion of making such an ultimatum. No confidential peer feedback system can function if there is a possibility of reviewers being de-anonymized at the demand of sufficiently influential coworkers. On a more opinionated note - making a demand like that displays critical lack of judgement in the important responsibility that all employees in supervisory roles have in ensuring a non-hostile working environment. Even if her intention was not to retaliate against her peers after determining their identities, it’s essential that people in positions of power do not take actions that even suggest that they are seeking to retaliate against less powerful peers. The importance of this principle is extreme to a healthy collaborative work environment. The fact that someone as intelligent as Gebru was surprised to be fired for making an ultimatum like this is itself surprising.

danielheath 5y ago

Sabotage is a pretty wild way to say “refusing to retract an academic paper that points out obvious flaws in ML models”.

andyxor 5y ago
there was much more to that, she demanded a list of reviewers who rejected her paper in a sort of ultimatum, and blasted group emails calling for sabotage I don't know if you saw Jeff Dean's response, but here it is https://docs.google.com/document/d/1f2kYWDXwhzYnq8ebVtuk9CqQqz7ScqxhSIxeYGrWjK0/preview?pru=AAABdlOOKBs*gTzLnuI53B2IS2BISVcgAQ https://docs.google.com/document/d/1f2kYWDXwhzYnq8ebVtuk9CqQ... There was also a thread on reddit with her colleagues reactions https://www.reddit.com/r/MachineLearning/comments/k77sxz/d_timnit_gebru_and_google_megathread/?sort=top https://www.reddit.com/r/MachineLearning/comments/k77sxz/d_t...
dekhn 5y ago
you left out "publicly calling out your SVP on twitter for not being racially sensitive enough" which she did before the paper blowup. Pretty sure that was the prime impetus for her being fired. BTW, I worked with TPUs on the hardware side, and the entire section about hardware power usage is just completely and totally wrong (I would have asked for the paper to be retracted if it was published, the errors are so large). As for the obvious flaws part, yeah, some of the flaws are obvious (and known to all practitioners of the art already), some are irrelevant but made to sound like they are existential crises, and some of the "flaws" are just per personal opionion about what she doesn't like. Hopefully she can find some academic position from which to propound her opinions. However, I do think in her behavior, she excluced herself from future employment.
- cdprojektdead 5y ago
  Multiple other, even white, employees called out this same SVP. Your usage of the phrase "Not being racially sensitive enough" strikes me as a bizarre euphemism for "racism", which it would appear you don't see as an issue. "are just per personal opinion about what she doesn't like." The field of AI Ethics is an amalgam of computer science and moral philosophy. What do you think value judgements are? What are those "personal opinions" and things she "doesn't like" and why? It seems you're using a lot of euphemisms to hide some very unpleasant truths about your world view.
  - 5y ago