Google finally fixes Gemini’s AI image generator – it went overboard with racial inclusivity

Google will soon return users’ access to the image generator in the Gemini AI chatbot. The feature was removed from the chatbot in February due to serious historical errors in its portrayal of people related to race and gender issues. For example, the racial diversity of soldiers for the query “Roman legion” is a clear anachronism.

Image source: Google

Early access to Google’s new Imagen 3 image generator will open to paying Gemini users on Advanced, Business and Enterprise plans in the coming days, Dave Citron, senior director of product at Gemini, announced on the official Google blog. Initially, the feature will only support queries in English.

«We have made technical fixes to the product, and also worked on more advanced evaluation algorithms and protection against red-teaming attacks,” Citron wrote.

In February this year, Google suspended Gemini’s image generation feature, citing that it offered “inaccuracies” in generating historical images. The company decided to send the Gemini image generator for improvement less than 24 hours after receiving the first complaints.

According to Citron, the new Imagen 3 generator “does not support the creation of photorealistic identifiable faces, images of minors, or overly gory, violent or sexual scenes.”

«Of course, as with any generative AI tool, not every image Gemini produces will be perfect, but we will continue to listen to user feedback and improve our product.”

He also promised that in the future more people will be able to use the image generator, and the function itself will receive support for additional languages.