Google's CEO reflected in a new interview on the company's Gemini AI image-generation debacle earlier this year, a controversy that drew backlash.
The AI assistant was quickly mocked for its apparent reluctance to generate images of white people and recreating images of historical figures with inaccurate ethnicities and genders, like Asian Nazis and Black founding fathers. Some critics used it as an example of "woke AI."
Google CEO Sundar Pichai told Bloomberg in an interview published Wednesday that "we got it wrong" and it was a case in which good intentions went awry.
Pichai said people from all over the world ask generic inquiries like "show me images of school teachers or doctors" and Google was trying to accommodate a global user base.
Advertisem*nt
"Obviously, the mistake was that we overapplied," Pichai said. "Including cases where it should've never applied so that was the bug."
Following criticism of the issue, Google paused Gemini from generating AI images of people while it corrected the changes.
Related stories
The issue hasn't been fixed yet, and if you ask Gemini to generate a picture of the founding fathers, it responds, "We are working to improve Gemini's ability to generate images of people" and that it expects the feature to return soon.
In February, Google DeepMind CEO said the image generator would return in a couple of weeks. Google didn't provide an update on when the feature will be readded.
Advertisem*nt
The Gemini chatbot has also faced some criticism in the past. Gemini has said it wouldn't promote meat or fossil fuels. Users on X have also complained that the chatbot inserts the word "diverse" into responses that don't call for it.
"We are rightfully held to a high bar and I think we clearly take responsibility for it," Pichai said in the interview. "And we are going to get it right."
Are you a Google employee? We want to hear from you. Email the reporter from a non-work device at [email protected]
