Google has announced that it is temporarily halting the creation of images with images of human figures in its advanced AI program Gemini. This decision allows the tech giant to refine the model for greater historical accuracy of its results.
Through an announcement on social media site X, the company described the move as a temporary suspension, noting its efforts to address “recent issues” related to historical inaccuracies in the images produced. “For now, we are suspending the creation of images featuring individuals and will soon introduce an enhanced version,” it said in a statement.
Google’s Gemini imaging tool debuted earlier this month, but it soon faced criticism for creating inappropriate images of historical figures that were widely shared on social media. For example, it portrayed the founding fathers of the United States in a way that was inconsistent with their actual ethnic origins, leading to backlash and ridicule.
Michael Jackson, a Paris-based venture capitalist, took to LinkedIn today to voice his criticism, calling Google’s AI efforts a misguided attempt at diversity, equity, and inclusion (DEI). Google previously admitted on X that the AI had inaccuracies when creating historical images, saying: “We are committed to improving these images immediately. While the Gemini AI does produce a variety of images of people, which is useful as it is used around the world, it has suffered from a lack of accuracy.”
Generative AI tools generate results based on their training data and other factors, such as model features. These instruments have been thoroughly tested for biased results, including overly sexualized images of women or images that imply racial stereotypes in occupational roles. A major controversy arose in 2015 when Google’s early AI technology incorrectly labeled black people as gorillas. Although the company pledged to fix the problem, a report from Wired years later revealed that the solution was just a temporary stopgap that prevented the technology from fully recognizing the gorillas.