- Google pauses Gemini’s ability to generate images of people after users complain about historical inaccuracies.
- Despite showcasing many new features at its I/O developer conference, Google has yet to fix Gemini’s image generation.
- Google is struggling to find a balance in addressing the issue due to biases in the training datasets.
Google faces challenges in fixing its Gemini AI image generator, which has been paused since February due to user complaints about historical inaccuracies and racial stereotypes. Despite promises for a quick resolution, the issue remains unresolved, highlighting the complexity of addressing bias in AI training datasets.
Gemini feature paused
Google’s Gemini AI image generator has been paused since February after users complained about inaccuracies in historical depictions and racial stereotypes. The AI, which was supposed to generate images of people, faced backlash for showing anachronistic and stereotypical representations.
Google CEO Sundar Pichai apologised, and Demis Hassabis of DeepMind promised a fix within weeks. However, the issue remains unresolved, and the feature continues to be disabled in Gemini apps on web and mobile.
Also read: Google is adding its Gemini Nano AI model to desktop Chrome
Fix faces challenges
The delay in addressing the problem highlights the complexity of dealing with biases in AI. Training datasets often contain more images of white people and reinforce negative stereotypes of non-white people.
Google’s initial attempt to correct these biases through hardcoding was clumsy, and now the company struggles to find a balanced solution. This ongoing issue serves as a reminder that correcting AI bias is a challenging task.
Also read: Google and HP to launch 3D video conferencing platform Project Starline
New features showcase
Despite the image generation setback, Google showcases other features of Gemini at its annual I/O developer conference. These include custom chatbots, a vacation itinerary planner, and integrations with Google Calendar, Keep, and YouTube Music. However, the unresolved image generation problem overshadows these advancements, underscoring the importance of addressing AI biases effectively.
