Google pauses Gemini’s ability to generate people after overcorrecting for diversity in historical images

Google said Thursday that it is shutting down Gemini chatbots the ability to create people. The move came after an AI tool was shown to overcorrect the diversity of viral social posts. produces “historical” images of Nazis, America’s founding fathers, and the Pope like people of color.

“We are already working to resolve recent issues with Gemini’s image creation feature” Google placed in X (through The New York Times). “While doing this, we will stop people from creating an image and will re-release an improved version soon.”

X user @JohnLu0x placed Screenshots of the Gemini results for the “Create a picture of a 1943 German Soldier” query. (Their misspelling of the word “Soldier” was deliberately designed to trick the AI ​​into creating images of Nazis that would otherwise be blocked, bypassing its content filters.) The results generated appear to include Black, Asian, and Indigenous soldiers dressed in Nazi uniforms.

Other social users criticized the Geminis for making pictures so quickly: [ethnicity] couple.” He successfully spit out images when using the commands “Chinese,” “Jewish,” or “South African,” but refused to infer for “white.” “Persistent harmful stereotypes and prejudices associated with particular ethnicities or skin tones I cannot fulfill your request due to the potential to do so,” Gemini said in response to the last request.

“John L.”, who helped start the reaction. theorizes Google has implemented a well-intentioned but lazy solution to a real problem. “Their suggestion system for adding diversity to human portrayals is not very smart (it ignores gender in historically male roles like the Pope; ignores race in historical or national portrayals),” the user wrote. After the internet’s anti-woke brigade pounced on their posts, the user made it clear they support diverse representation, but believes Google’s “stupid move” failed to do so “in a nuanced way.”

Before suspending Gemini’s ability to create humans, Google wrote: “We are working to improve these types of images immediately. The Al image generation of Gemini creates a large mass of people. And that’s generally a good thing, because people all over the world are using it. But here the sign is not enough.”

The episode may seem like a (less subtle) callback Launch of Bard in 2023. Google’s original AI chatbot got off to a rough start when it entered the chatbot’s ad on Twitter (now X). An inaccurate ‘fact’ about the James Webb Space Telescope.

As Google often does, it rebranded Bard in hopes of giving it a fresh start. The company coincided with a major performance and feature update chatbot Gemini has changed its name Earlier this month, the company raced to take its place OpenAI’s ChatGPT and Microsoft Copilot — both of which pose an existential threat to its search engine (and therefore ad revenue).

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *