Google has temporarily suspended the ability to render images of people in its conversation app Gemini, citing improvements needed to ensure accuracy. This decision was made following user feedback pointing to errors and disturbing results in the rendered images. This feature, based on the Imagen 2 AI model, encountered issues in creating relevant and unbiased representations, leading to its suspension.

Google will improve artificial intelligence for respectful and accurate representation

Google acknowledges these issues and attributes them to difficulties fine-tuning its AI model to avoid inappropriate or biased representations while striving for inclusivity. Despite efforts to alleviate problems seen with previous imaging technologies, this feature has faced significant hurdles. Sometimes it created unnecessary images or was very careful not to create images for certain prompts.

Senior Vice President Prabhakar Raghavan explained that the goal is to resolve these issues with significant improvements and testing before making the feature available again. He also highlighted ongoing efforts to improve the accuracy of the technology, noting the challenges associated with ensuring the reliability of AI, especially on sensitive issues.

Even public figures, including Elon Musk and Republican leader Vivek Ramaswamy, have criticized the feature for creating images that they say are historically inaccurate and racially insensitive.

The suspension is part of Google’s broader efforts to develop AI technologies responsibly, taking into account the complexities involved in building creative and accurate AI systems. The company said it is working diligently to resolve recent issues and plans to re-release an improved version of the rendering feature soon.

Google also recommends that you use Google Search to obtain the most up-to-date and reliable information from a wide variety of web sources through different systems.