AFP Paris

Paris

Updated Thursday, February 22, 2024-13:32

  • Technology Google Bard is dead, long live Gemini

Google

announced this Thursday that it would suspend the creation of images of people in its generative artificial intelligence (AI) tool

Gemini

, after "recent problems" with this functionality.

The computer giant launched this new software in

the United States

at the beginning of February , which is not yet available in

Europe

.

Since then, criticism has emerged on social media to point out inaccuracies regarding certain images created, particularly in terms of genre and diversity of historical themes.

In this case, users noticed on X (formerly Twitter) that Gemini appeared to underrepresent white people in the generated images.

Among them, several Internet users published a request for a German soldier from 1943, resulting in images of Asian or black-skinned soldiers.

"We are working to resolve recent issues with Gemini's imaging feature. In the meantime, we will be pausing people imaging and will have an enhanced version of this feature available soon," said

Jack Krawczyk,

Google's product manager in charge. of Geminis, in a statement sent to AFP.

"Gemini's AI imaging results in a wide variety of people. And that's usually a good thing because people are using it all over the world. But in this case we lost it," Google insisted on X.

In a reaction published on this same social network, of which he is the owner,

Elon Musk

published an image where he distinguishes between, on the one hand, his own AI tool project, assimilated to "the ultimate search for truth" .

", and on the other hand Gemini and the creator of ChatGPT,

OpenAI

, whom he accuses of being "racist."

Since the end of 2022 and the success of ChatGPT, generative AI, capable of producing all types of content (texts, sounds, images or videos) from a simple request in everyday language, has aroused enormous enthusiasm and all the giants Technologists are engaged in a race to implement tools for organizations and individuals.

This increasingly surprising technology, right down to the ability to create one-minute videos with OpenAI's

Sora software

introduced in mid-February, has also led to safeguards being put in place.

Last week, twenty of the most advanced companies in this field, including Meta, Microsoft, Google and OpenAI, announced their commitment to developing new techniques to identify disinformation content using AI.