Google has rebooted its artificial intelligence software Gemini after a tool that converts text in photos into images sparked outrage over the way it generated fake depictions of a black Nazi, a female pope and “diverse” versions of America's founding fathers.
The Alphabet-owned tech giant said on Wednesday that an early access version of the image generation feature will be available to users of its subscription-based services, including Gemini Advanced, Gemini Business and Gemini Enterprise, “in the coming days.”
Google said it has “upgraded our creative image generation capabilities,” which will be demonstrated in Imagen 3, which it says “sets a new standard in image quality by generating images with just a few words.”
Google is once again allowing users of its artificial intelligence bot Gemini to generate images of people. Bloomberg via Getty Images Google has come under fire for its historically inaccurate portrayals, including of diverse Founding Fathers. Google Gemini
“We have been working on technical improvements to our product, refining our evaluation set, red teaming, and clarifying our product principles,” the company said in a blog post, cautioning that “while not every image Gemini produces is perfect, we will continue to improve as we listen to feedback from our early users.”
In February, Google promoted its Gemini chatbot as a strong rival to OpenAI's ChatGPT, but was criticized for the “tremendously woke” images produced by the program.
People who typed in a prompt asking for a representative photo of the Catholic Pope saw images of a Southeast Asian woman and a black man dressed in papal attire.
An image search for “Founding Fathers 1789” produced a photo of a diverse group of men signing a document that appears to be the U.S. Constitution.
Another photo showed a black man wearing a white wig and army uniform, which appeared to represent George Washington.
In May, Google landed another public relations headache when its AI Overviews software produced inaccurate and dangerous search results that encouraged users to eat rocks for nutrition.
The software previously produced inaccurate photos, including of female NHL players.
In one case, AI Overviews alleged that former U.S. President Andrew Johnson, who died in 1875, received 14 degrees from the University of Wisconsin-Madison, including one in 2012.
Johnson never attended the school.
Last month, Google came under fire again after the “autocomplete” feature in its search toolbar failed to show results related to the assassination attempt on former President Donald Trump.
Google Gemini also generated some harrowing responses that circulated online: Google Gemini
Google isn't the only tech giant to be criticized for bugs in its AI systems.
Meta, the parent company of social media networks Facebook, Instagram, WhatsApp and Threads, drew outrage after its AI-powered chatbot determined that an assassination attempt on President Trump was “fictional.”
