Why was Google's AI Ethics Board unsuccessful

Google comes under fire for firing a well-known AI researcher

Google hit the headlines just two days ago after a complaint from the US labor protection agency became public. According to the letter from the National Labor Relations Board (NLRB), two employees were wrongly fired at the end of 2019 for their activist efforts after organizing employee protests against the Group's collaboration with the consulting firm IRI Consultants. Now Timnit Gebru, a senior AI ethics researcher, has been accused of being fired because of an email sent to her colleagues.

Rejected research paper

In the said letter, Gebru expressed her displeasure that she had been asked to withdraw a research paper. "A week before you go on vacation, an appointment will appear in your calendar. In the meeting, your supervisor will tell you that it has been decided that you have to withdraw the research paper by next week. You are not worth a conversation because of your humanity and expertise is not recognized or valued by this company ", one reads in their message, which was published in full by" Platformer ".

As a result, Gebru informed her manager that she met certain requirements for her work in the group and that she would respond as soon as she returned from vacation. In response, Gebru received a notification that her resignation was accepted - although she did not want to resign, as she announced on Twitter.

In another posting on the short message service, she also alleges that the decision to terminate her employment relationship with immediate effect was made by Jeff Dean, the head of the AI ​​department at Google.

Two perspectives

On Thursday, Dean tried to explain the situation to Google employees from a company perspective. "Timnit co-authored a research paper with other Google employees and outside collaborators. As always, submitted research reports go through a review process that takes two weeks. Unfortunately, this specific work was submitted one day before the agreed deadline and - instead of feedback to wait - cleared for submission, "writes Dean. The authors were then informed that the publication did not meet the quality requirements and why that was the case.

Timnit Gebru is known for her work on "algorithmic bias". Together with Joy Buolamwini, the founder of the Algorithmic Justice League, she showed in a research paper in 2018 that the error rates in recognizing people of color are far higher than with whites, as "The Verge" reports. It remains to be seen whether the National Labor Relations Board will file a complaint in this case as well. (mick, December 4th, 2020)