Google’s AI-powered image generation tool, Gemini, recently came under fire for inaccurately depicting historical events and figures, drawing widespread criticism and ridicule, particularly on social media platforms like Twitter. This led Google to temporarily suspend the tool’s image creation feature. Sundar Pichai, Google’s CEO, has since acknowledged the shortcomings of Gemini in an internal memo leaked by Semafor, admitting the tool’s errors were “unacceptable” and committing to ongoing improvements, despite acknowledging that no AI system is flawless.

Originally named Bard, Gemini’s image generation function was paused after users reported it produced historically inaccurate images, such as portraying Vikings and the Founding Fathers as people of color. This sparked debates over “woke” programming. Google’s Senior Vice President, Prabhakar Raghavan, clarified that the inaccuracies were unintentional, resulting from an overemphasis on diversity without proper context. He also noted that Gemini had become overly cautious, sometimes rejecting reasonable image requests to avoid potential sensitivities, including refusing to generate images of white individuals in some instances.

Google plans to reintroduce Gemini’s image generation feature with comprehensive updates to address these issues, including structural changes, revised guidelines, enhanced launch protocols, and improved evaluation and testing processes.

Pichai has expressed regret over the AI’s problematic outputs related to racial bias and other issues, emphasizing the company’s commitment to rectifying these errors. The controversy has raised broader concerns about the risks of AI bias as AI technology becomes more integrated into daily life. Criticism has also come from notable figures like Tesla’s Elon Musk, who accused Google of promoting “racist, anti-civilizational programming” through Gemini.

Despite these challenges, Google is determined to refine Gemini, with Pichai stressing the importance of learning from these mistakes and ensuring they are corrected at scale. This incident underscores the complexities and responsibilities involved in developing AI technologies, especially as they gain the ability to shape perceptions and narratives.