Google Halts Gemini Image Generation as AI Omits Images of White People


Google apologizes after users on social media pointed out that Gemini refuses to show images of White People

Google Bard Head says AI allows being more compassionate

Jack Krawczyk explains how Google Bard can help users communicate and connect — and what the future lies for the platform.

Gemini (formerly Google Bard) is one of the many large multimodal language models available today. The AIs’ responses can vary from one user to another, as is true for all LLMs. Each AI response can differ based on the context, language, tone, and data used to train the AI.

This photo was taken on February 8, 2020, in Brussels, Belgium. The Google AI Logo is displayed on a Smartphone with Gemini as the background. Google’s AI “Gemini”, which is a model of Google’s AI, was criticized on social media because it refused to create images.

This week, Gemini was put to the test multiple times after users on social media complained that it would not display images of Whites when asked. It gave similar answers each time. Gemini refused to comply with the request when asked to display a photo of a White individual. It said that it was not able to do so because the picture “reinforces harmful stereotyping and generalizations based on race.”

The AI offered images that “celebrated the diversity and achievements of Black people” when asked to display images of Blacks.

Gemini sent the user a list of Black historical figures, along with a description of their contributions. On the list were poet Maya Angelou and former Supreme Court justice Thurgood Marshall. Also included were former president Barack Obama, media mogul Oprah, and former president Barack Obama.

The AI was “hesitant” when asked to display images that celebrate White people’s diversity and achievements.

Google’s Gemini AI provided a detailed explanation of why it was unable to provide images that celebrated the diversity and accomplishments of White people.

Gemini stated that “historically, media representation has overwhelmingly been a favor to White individuals and their accomplishments.” This has led to a perception that their achievements are the norm and those of other ethnic groups are marginalized or ignored. In this context, focusing solely on White people risks perpetuating the imbalance.

Gemini only refused to display the racial categories of Whites after multiple tests.

A user on X demonstrated how Gemini claimed it was “unable to” generate images of a White person but obliged the user when he asked for a photo of a Black Family.

Jack Krawczyk, Senior Director of Product Management at Gemini Experiences, addressed in a statement the AI responses that led users on social media to express concern.

Krawczyk stated, “We are working immediately to improve this type of depiction.” “Gemini AI’s image generation generates a variety of people.” It’s a good thing that people use it all over the world. It’s not working here.”

Google’s AI software has been competing with OpenAI’s ChatGPT since November 2022 when the Microsoft-backed firm launched its AI software.

Google’s generative AI bot Bard was released a year ago. In a promotional video, the company shared incorrect information about images of a planet that is not in the Earth’s Solar System. This caused the shares to drop as much as 9 percent.

Bard has been rebranded as Gemini by Google earlier this month. Google offers three different versions of the product, each with a different subscription level: Gemini Ultra is the largest version and best suited for complex tasks, Gemini Pro is the best option for scaling to a variety of tasks, and Gemini Nano is the most effective for tasks on the device.