The tool had been singled out for depicting Vikings. (new window) and soldiers of Nazi Germany as black or Asian people.
Google’s Gemini artificial intelligence had depicted Vikings as racialized people.
Photo : Gemini
At the time, a Google spokesperson stressed that the fact that Gemini generates a wide range of people was generally a good thing
given the diversity of people using the tool around the world. He did, however, concede that he missed his shot in this case
.
The functionality was therefore suspended while it was being improved.
Six months later, it’s done: Google claims that the technical capabilities of its model, named Image 3offer better performance.
For now, image generation on Gemini will be available, in English only, to users with the paid version of Gemini. The model will be extended to a larger number of people and languages later.
Like most mainstream generative models, the tool does not support creating images of children, identifiable people, or sexually explicit or violent content.
The arrival of the Gems
The Californian giant took advantage of the reinstatement of the generation of images of people to announce the deployment, in the coming days, of its Gems functionality, its new personalized conversational robots, (new window) according to specific instructions. OpenAI has been offering a similar feature for several months, which it calls GPTs. (new window)
You can create any Gem you dream of: a gym buddy, a sous chef, a coding partner, or a creative writing guide. They’re easy to set up, too
Google described in a blog post in May.
However, you must have an account Gemini Advanced, Gemini Business or Gemini Enterprise to access it.
