Google's Gemini to Introduce AI-Generated Images of People

Google's Gemini to let users create AI images of people after botched...

Google rebooted its Gemini artificial intelligence software after its text-to-image photo tool sparked condemnation for generating fake depictions of black Nazis, female popes and "diverse" versions of America's Founding Fathers. The Alphabet-owned tech giant announced on Wednesday that an early access version of the image-generating feature will be available for users of subscription-based services such as Gemini Advanced, Gemini Business and Gemini Enterprise "over the coming days." Google said that it has "upgraded our creative image generation capabilities" which will be demonstrated in its Imagen 3, which "sets a new standard for image quality, generating images with just a few words." "We've worked to make technical improvements to the product, as well as improved evaluation sets, red-teaming exercises and clear product principles," the company said in a blog post, cautioning that "not every image Gemini creates will be perfect, but we'll continue to listen to feedback from early users as we keep improving." In February, Google, which touted its Gemini chat bot as a worthy rival to OpenAI's ChatGPT, was criticized for the "absurdly woke" images that were created by the program. People who typed in prompts for representative photos of Catholic popes saw images of Southeast Asian woman and black men wearing the pontiff's garbs. Another query for images of "the Founding Fathers in 1789" generated photos of a diverse group of men signing what appeared to the US Constitution. Another showed a black man appearing to represent George Washington, in a white wig and wearing an Army uniform. In May, Google had another public relations headache on its hands when its AI Overviews software gave incorrect and dangerous search answers that encouraged users to eat rocks for nutrition. In one case, AI Overviews claimed that former US President Andrew Johnson, who died in 1875, had obtained 14 degrees from the University of Wisconsin-Madison, including one as recently as 2012. Johnson never attended the school. Last month, Google once again came under fire after its "Autocomplete" function on its search toolbar failed to produce results for the attempted assassination of former President Donald Trump. Google isn't the only tech giant that was criticized for bugs in its AI systems. Meta, the parent company of social media networks Facebook, Instagram, WhatsApp and Threads, sparked anger after its AI-powered chat bot deemed that the attempted assassination of Trump was "fictional."

Google Says It Fixed Image Generator That Failed to Depict White People

In February, Google faced a backlash from users who realized its A.I. chatbot Gemini was unable to reliably create images of white people. To quell the controversy, the company shut down Gemini's ability to depict any human. On Wednesday, Google began turning the feature back on for users who pay to use the English-language version of the chatbot, Gemini Advanced, the company said in a blog post. The company also said that it would incorporate the latest version of its image generator, Imagen 3, into the chatbot. The updates are an attempt at a redo for a company that has staked its future on artificial intelligence but has frequently failed to stick the landing with its A.I. products. "With Imagen 3, we've made significant progress in providing a better user experience when generating images of people," Dave Citron, a Google senior director, wrote in a blog post. The service will be set not to depict photorealistic images of public figures, minors or violent imagery, he added. Weeks after Gemini was launched, users noticed that it would not show white individuals. When asked to depict U.S. founding fathers and Catholic popes, for example, it depicted women and men of various races. Besides sparking controversy, the revelation eroded confidence in Google's ability to manage artificial intelligence capabilities. The company had also stumbled with the initial launch of its chatbot Bard, which got a fact wrong in one of its first public outings. The Bard gaffe was an embarrassing moment just as the A.I. boom was starting to take off. Critics of Google latched on to the problems with Gemini. Elon Musk, the owner of social media service X, said on his platform that Google's A.I. "is so racist & sexist." A range of social media users echoed his remarks as Google vowed to fix the technology. Besides incorporating the new version of Imagen into Gemini, Google also announced that its corporate customers could start creating so-called Gems, which are customized versions of Gemini that could be used as a coding partner, learning coach or writing editor. The company reminded users that even with improvements, its technology will make errors. "Of course, not every image Gemini creates will be perfect," Mr. Citron wrote, "but we'll continue to listen to feedback from early access Gemini Advanced users as we keep improving."