Google Temporarily Suspends People Image Generation on Gemini

Google has temporarily suspended the people image generation feature in its Gemini conversational app due to concerns about accuracy and potential offensive outputs. The decision comes in response to user feedback highlighting inaccuracies and offensive representations in the generated images. Powered by the AI model Imagen 2, the feature struggled to produce appropriate and unbiased depictions, leading to its suspension.

Acknowledging these challenges, Google cites difficulties in fine-tuning the AI model to ensure respectful and accurate representations while striving for inclusivity. Despite efforts to address issues encountered in previous image-generation technologies, the feature faced significant hurdles, sometimes producing unnecessary or overly cautious depictions.

Senior Vice President Prabhakar Raghavan emphasizes the need for extensive improvements and testing before reintroducing the feature. He acknowledges the complexities of ensuring AI reliability, especially concerning sensitive topics, and underscores ongoing efforts to enhance accuracy.