> Google has terminated the head researcher on its AI ethics team over what the tech giant deemed a violation of its “code of conduct,” as the company faces an internal revolt over a previous firing of a black data scientist. Margaret Mitchell, who led Google’s Ethical AI unit, announced her termination in a brief tweet on Friday, saying only “I’m fired” without offering further detail.
> Google’s own statement on the matter was also somewhat vague, confirming Mitchell was let go over “multiple violations of our code of conduct,” while adding that she ran afoul of “security policies” by downloading “confidential business-sensitive documents and private data of other employees.”
> The move comes after the AI team’s co-head, Timnit Gebru, was fired in December during a dispute over a research paper she authored that was critical of large-scale AI language models, which are used in many Google products. The company demanded she remove her name from the work and fired her when she refused. At the time, Gebru said she had offered a number of conditions on which she would take her name off the research paper, but was forced out regardless.
> Both Mitchell and Gebru have also taken Google to task for an alleged lack of diversity among its staff, with Gebru writing in an internal email just before her firing that “your life gets worse when you start advocating for underrepresented people” at the company, and that she’d been “constantly dehumanized” by superiors.
> Mitchell was highly critical of Gebru’s ouster, penning a lengthy message to the company last month voicing concerns about its “relation to sexism and discrimination,” also noting she had been “cut off” from Google’s internal messaging system. According to a source cited by Axios, the lock-out came after she scoured company correspondence for evidence of Gebru’s claims of harassment and discrimination, consistent with Google’s statement on Friday about her unauthorized access to “confidential” documents.
[–]Gave 1 insightful - 1 fun1 insightful - 0 fun2 insightful - 0 fun2 insightful - 1 fun - (0 children)