Ex-Google scientist Gebru opens AI institute year after tumultuous exit

Timnit Gebru, the computer scientist whose disputed exit from Google's artificial intelligence research team prompted debate across the tech industry about diversity and censorship, said on Thursday she has launched a small lab to continue her work freely.

The Distributed AI Research Institute has raised US$3.7 million from foundations and aims to critically study services from big tech companies as well as propose AI-based solutions to issues such as food insecurity and climate change, Gebru said.

It joins several non-governmental projects such as the Algorithmic Justice League that are advancing ethical use of AI. Critics worry that without proper safeguards systems including for facial recognition and credit scoring could lead to mass surveillance and racial discrimination.

Gebru has hired a fellow based in South Africa and expects to add other researchers next year. They will publish studies and educate activists and lawmakers globally.

"I want to make sure DAIR is not just working on research papers," Gebru said. "I want to be an institute that realizes you have to engage with various groups of people."

Gebru, who is Black, has said Google fired her a year ago for criticizing its lack of workforce diversity and for fighting managers who objected to publishing a paper she co-wrote on potential social and environmental costs of language technology. Google has said it accepted Gebru's resignation.

Her speaking out about the incident drew praise from many scientists and engineers, but others questioned her work and tactics. Alphabet Inc unit Google in the aftermath reorganized the ethical AI research team Gebru had led, fired her co-leader and lost the pair's manager to Apple.

Freedom to pursue whatever led Gebru to start DAIR over joining another company. But sustaining it without becoming beholden to sponsors or other powers will be the challenge, she said. Initial backers include the MacArthur, Ford and Rockefeller foundations.