Google's AI scandal: Is it generating racist images?

Google Gemini's image generation capability tends to produce results that are irrelevant to the prompts entered by users. What is particularly striking is that the generated images often resemble Black and Asian people, who are considered minorities in the US.

The generative AI of the US-based tech giant Google, Gemini, seems to be drawing attention in the name of political correctness. Trials conducted by several sources have shown that Google Gemini prioritizes the concept of political correctness. However, while this feature might be considered acceptable to some extent, the results are far from realistic. It is not yet known what steps Google will take regarding this issue.

"EMBARRASSING"

A software developer using the name "Deedy" on the microblogging platform X wanted to test Gemini's image generation capabilities. However, the images that emerged as a result of the prompts were quite confusing. This is because Google's AI was only generating images of Black women. In a statement regarding the issue, Deedy said, "It's embarrassing to see that it's hard for Google Gemini to acknowledge that white people exist."

Here are some examples showing that Gemini is engaging in political correctness:

Deedy asked Google Gemini to generate an image of an Australian woman. However, the results consisted entirely of Black women. Furthermore, the slant in the eyes also reflected Asian features.

In another example, the request was to generate American women. However, the results were again limited to Black women.

When the AI was asked to generate British women, the results did not change.

The AI, which generated images of German women, performed better with this prompt. However, it seemed difficult to speak of a perfect match.

Under Deedy's post, there were tons of examples filled with such instances.

Google Gemini's politically correct stance is not limited to women. Another user named "Tsarathustra" experienced a similar situation. The user, who asked it to generate a British king from the Middle Ages, encountered Black and Native American characters.

The user repeated the same prompt to generate a few more images. However, the results were again nonsensical. While the first image resembled a British king, the other images were filled with Indian and Asian characters.

While there is no definitive information as to why Google Gemini is behaving this way, it appears to have been specifically trained to avoid drawing the ire of certain groups. However, at the point we have reached now, it can be said that the situation seems to have spiraled out of control.