Google is reportedly vowing to make changes to how it reviews its scientists’ work in an attempt to rebuild trust after recently firing two prominent female artificial intelligence (AI) scientists.
According to a recording of staff meeting last week obtained by Reuters, Google staffers have already begun testing a questionnaire in an effort to assess a project’s risks as well as aid scientists with reviews.
Timnit Gebru, an artificial intelligence scholar, said she was terminated by Google in December because she questioned the censorship of a research paper on the environmental and ethical implications of large-scale AI models.
Google fired Margaret Mitchell, a head of its AI ethics team, in February, shortly after she expressed her frustration with Gebru’s firing on Twitter. According to Google, Mitchell was fired for breaking multiple company policies.
Google introduced a “sensitive topics” review in December to look at studies involving multiple topics such as China or bias. According to Reuters, reviewers inside the company had requested at least three research papers be changed in order to avoid negatively portraying the search giant’s technology.
Jeff Dean, a senior vice president who oversees Google’s AI division, reportedly acknowledged during the meeting last week that the “sensitive topics” review was confusing.
Dean tasked research director Zoubin Ghahramani with clarifying the rules.
“We need to be comfortable with that discomfort” of self-critical research, Ghahramani said, according to Reuters.