Google Apologizes After Gemini AI Generates Historically Inaccurate Images
Google admits its Gemini AI image generator "missed the mark" after facing severe backlash for creating historically inaccurate and offensive images. The company pauses the feature while promising to fix the overcorrected diversity filters.
Google issues an apology after its Gemini AI image generator faces intense criticism for producing historically inaccurate images. Users discover that the tool creates "woke" content by inappropriately inserting diverse ethnicities into historical contexts while often ignoring White people entirely. A top Google executive openly admits that the generated images are inaccurate and offensive, acknowledging that the feature completely misses the mark.
Prabhakar Raghavan, a senior vice president at Google, publishes a blog post explaining exactly what goes wrong with the AI. He states that the system is originally tuned to avoid violent or sexually explicit content and to show a range of people rather than a single ethnicity. However, he admits that these safety filters fail to account for situations where showing a diverse range of people is historically inaccurate or inappropriate.
In response to the widespread backlash, Google pauses the image generation of people entirely. The company pledges to use the recent feedback to improve the AI's understanding of historical and cultural contexts. Raghavan emphasizes that when users ask for a specific type of person or a particular historical setting, the AI absolutely needs to provide an accurate reflection of that specific request.