Google CEO Sundar Pichai is not happy about what’s been going on with the company’s Gemini AI chatbot, calling viral and historically inaccurate errors from its image generator “unacceptable” in an internal memo to employees on Tuesday.
“I want to address the recent issues with problematic text and image responses in the Gemini app (formerly Bard),” Pichai said in the memo, which was first reported by Semafor and later confirmed by The Verge. “I know that some of its responses have offended our users and shown bias—to be clear, that’s completely unacceptable and we got it wrong.”
Pichai explained that teams at Google were “working around the clock” to address the issues in Gemini and had “seen substantial improvement on a wide range of prompts.” Gemini’s image generator caused an online firestorm last week after the AI chatbot created historically inaccurate images of Black Vikings and refused to generate images of white people.
Google quickly paused Gemini’s ability to create images of people last Thursday in light of the scandal, with Google vice president Prabhakar Raghavan apologizing for the mishap on Friday.
“This wasn’t what we intended. We did not want Gemini to refuse to create images of any particular group. And we did not want it to create inaccurate historical—or any other—images,” Raghavan said in a blog post.
The company aims to relaunch Gemini’s image generator over the next few weeks.
While Raghavan’s statement was meant for the public, Pichai’s memo was addressed to employees, some of whom may have been taken aback by Gemini’s outputs. One person who claims to work at Google’s AR & VR unit in San Francisco said they had “never been so embarrassed to work for a company” when Gemini’s historically inaccurate images of a medieval British king went viral. Gizmodo has not been able to independently confirm whether this person works at Google.
In his memo, Pichai said the company would make structural changes to ensure the situation with Gemini didn’t repeat itself. He also stressed that it was important for Google to give users “helpful, accurate, and unbiased” information about all its products, including its emerging AI products.
“No AI is perfect, especially at this emerging stage of the industry’s development, but we know the bar is high for us and we will keep at it for however long it takes. And we’ll review what happened and make sure we fix it at scale,” Pichai said.