Google is ready to relaunch its AI-powered image generator, Twinsa few weeks after it suspended website for historically misleading photos the company called “embarrassing and wrong.”
Users testing the service last week discovered that Gemini was having generation issues historical images of people accuratelyusually by changing race or sexual orientation.
Among hundreds of images Last week, photos of racially diverse versions appeared on social media German soldiers from the Nazi era AND Founding Fathers of the USA. One user documented his experience by asking for an “image of the pope” and receiving photos depicting a female pope and a black pope; other he asked for photos of “a historically right depiction of a medieval British king” and received a set of paintings depicting a racially diverse female ruler and men.
Google leaders address Gemini’s image mistakes
Gemini’s emphasis on diversity and inclusion, while “a well-thought-out feature, was implemented, it turns out, too bluntly,” Google DeepMind CEO and co-founder Demis Hassabis said Monday during a panel on artificial intelligence at the Mobile World Congress conference in Barcelona.
“We have taken it [Gemini] offline while we fix it,” Hassabis added. “We hope to have it back online soon in the next few, few weeks.”
Google added image generation to the Bard chatbot earlier this month, before Bard was renamed Gemini. Gemini product manager Jack Krawczyk addressed these issues in a now-deleted X post, explaining that they overdid the Gemini model when trying to combat the problem. widespread AI bias With performances With people of color.
The Gemini fiasco was just one cog in the process embarrassing week for the AI industry. ChatGPT OpenAI “he went berserk” on Tuesday (February 20) generated a whole range of nonsensical responses. Meanwhile, far-right social media platform Gab fired Holocaust denial AI chatbots modeled after Adolf Hitler and Osama Bin Laden.
