Google has apologized for what it describes as “inaccuracies in some historical images” created using the Gemini AI tool. The wide range of results that Gemini sought to achieve in the generation of images turned out to be too wide in fact – politically correct AI began to show racial diversity where it did not exist and a priori could not be – for example, in pictures of German soldiers in 1943, or in portraits of the founding fathers of the United States, among whom, as it happened, there was not a single African-American. This is not the first time that stunned users have drawn Google’s attention to the flaws in the work of Gemini. The answer was not long in coming.

We are aware that Gemini is somewhat inaccurate in some historical facts when generating images. We are working on immediate improvements to these images. Gemini generates images of a wide range of people. And that’s generally a good thing because it’s used by people all over the world. But here it definitely lacks quality.

In the meantime, Google is working on immediate improvements to Gemini’s algorithms, it simply refuses to generate some images. Google’s AI, for example, ignored reporters’ requests to create images of Vikings, Third Reich officials, and 19th-century U.S. presidents, apparently not knowing whether they included all races.

Літературний редактор.

Commentary