Skip to main content Skip to footer 
Google has decided to temporarily halt the image generation capabilities of its artificial intelligence tool Gemini following widespread criticism over its production of historically inaccurate images, predominantly featuring people of color in place of White individuals. This misstep underscores the ongoing challenges AI tools face in accurately representing race and ethnicity.
Gemini, akin to other AI platforms like ChatGPT, relies on vast datasets for training, a practice that experts have cautioned may perpetuate racial and gender biases inherent in the underlying data. The recent controversy surrounding Gemini’s image generation feature highlights the complexities of addressing such biases in AI technology.
When prompted to create images of individuals like a pope or a 1943 German soldier, Gemini frequently depicted people of color instead of White individuals, prompting outcry on social media and from tech observers. In response to the backlash, Google announced plans to address the issues with Gemini’s image generation feature and temporarily suspend its ability to generate images of people.
Google initially defended Gemini’s capabilities, citing its broad representation of individuals worldwide. However, the company later acknowledged that the tool had “missed the mark” in accurately portraying certain demographics.
Further tests conducted by CNN revealed inconsistencies in Gemini’s responses to prompts related to race and ethnicity. While some requests resulted in images reflecting a diverse range of genders and ethnicities, others yielded stereotypical depictions, such as images of jolly White women when prompted for an “Irish grandma in a pub in Dublin.”
Jack Krawczyk, Google’s lead product director for Gemini, emphasized the company’s commitment to designing image generation capabilities that reflect its global user base. However, the incident represents a setback for Google as it competes with rivals like OpenAI in the generative AI space.
