Timnit Gebru, the artificial intelligence (A.I.) researcher whose departure from Google set off a major controversy, is launching a new research institute committed to the responsible development of A.I.
The institute, Distributed Artificial Intelligence Research (DAIR), will take an inclusive approach to research, as well as hiring tech talent. “We believe that research should center the voices and experiences of those most impacted by technology and should be rooted in their communities,” reads a note on DAIR’s website. “Our distributed research institute allows us to remain embedded in our communities in the U.S. and globally to include a greater diversity of perspectives and lived experiences.”
Unlike many other A.I. research centers, DAIR also claims it will try to stop the development of A.I. that can potentially cause harm: “Our goal is to be proactive about this technology and identify ways to use it to people’s benefit where possible, caution against potential harms and block it when it creates more harm than good.”
In December 2020, Gebru, who was co-lead of Google’s Ethical Artificial Intelligence Team, publicly claimed that Google management asked her to remove her name from a paper suggesting that a large-scale A.I. language model (LLM) can generate biased results. She also claimed that, after she refused to do so, management fired her. (Google executives subsequently insisted that she resigned, which she denied.)
As the resulting controversy raged, Google CEO Sundar Pichai issued a public apology. “I’ve heard the reaction to Dr. Gebru’s departure loud and clear: it seeded doubts and led some in our community to question their place at Google,” he wrote in an internal memo. “I want to say how sorry I am for that, and I accept the responsibility of working to restore your trust.”
That failed to cool the situation, with prominent Google engineers quitting over the controversy. In the wake of those departures, Google announced plans to restructure its Ethical Artificial Intelligence (A.I.) Team.
Early DAIR projects will include the creation of a database of aerial imagery of South Africa “to examine how the legacy of apartheid is still etched into land use,” according to Wired. At the moment, DAIR is a fiscally sponsored project of Code for Science & Society (CS&S), but will eventually become its own nonprofit.