Google has announced it is temporarily halting its Gemini AI image generator after wide critics slammed the generative AI system for depicting “inaccuracies in some historical image generation depictions.” For context, Google’s Gemini AI began offering text-to-image generative features this month, but its attempts at creating a wide range of results have missed the mark.
Prior to Google halting its Gemini AI image generator, the tech giant had apologised on X saying : “We’re aware that Gemini is offering inaccuracies in some historical image generation depiction. We’re working to improve these kinds of depictions immediately. Gemini’s AI image generation does generate a wide range of people. And that’s generally a good thing because people around the world use it. But it’s missing the mark here.”
However, after bizarre results generated by Gemini AI, such as a racially diverse group of images depicting a 1943 German soldier and non-white AI-generated people in the Founding Fathers, Google immediately halted proceedings on its Gemini AI image generator and paused the feature until they resolved the issue.
Also Read: NVIDIA’s Financial Triumph Amidst the AI Boom and Collaboration with Google
Gemini AI Image Generator Is Intentionally Avoiding White People?
After Google paused its Gemini AI image generator feature due to inaccurate racial historical depiction, conspiracy theories online spread that Google is intentionally avoiding depicting white people. For example, people have found that Gemini AI has generated inconsistent images such as Black Vikings, a Black Catholic pope, only Black women NHL players and non-white people in a scene of the US Founding Fathers.
In an experiment done by The Verge, Gemini’s AI images were seen essentially erasing the history of race and gender discrimination. For instance, when they placed the prompt of a US senator from the 1800s, the results were Black and Native American women despite the first female senator in 1922 being White.
Moreover, earlier statements this week by former Google employees stated it is “embarrassingly hard to get Google Gemini AI to acknowledge that white people exist”, as the generative AI system overwhelmingly produces non-white AI-generated people as results from prompts such as historical groups or figures like the Founding Fathers, the conspiracy to avoid depicting white people is not without reasoning.
Computer scientist and investor Paul Graham blasted Google in his X post, “The ridiculous images generated by Gemini aren’t an anomaly. They’re a self-portrait of Google’s bureaucratic corporate culture.” Lastly, Right-wing commentator Ashley St. Clair has framed the controversy in a broader culture-war context. “What’s happening with Google’s woke Gemini AI has been happening for years in Media and Hollywood, and everybody who called it out was called racist and shunned from society.”
What Are Google Plans Moving Forward for Gemini AI?
Regarding Gemini AI’s image generation, Google has responded with a statement on X “We’re already working to address recent issues with Gemini’s image generation feature. While we do this, we’re going to pause the image generation of people and will re-release an improved version soon.”
As for the improvements, Google revealed that its image generation tool will be enhanced with its AI model, Imagen 2, allowing brands to create images featuring people. However, with what just happened, Google stated they would implement safety measures such as restrictions against generating known individuals, celebrities, public figures, or content depicting violence or illegal activities.
In addition to a revamped version of Gemini AI image generation, Google has stated its plans to enhance its advertising products by integrating Gemini AI models, aiming to offer advanced AI capabilities to a broader range of its customers. For instance, the tech giant plans to incorporate generative AI in its enterprise tools by producing text and realistic images.
Lastly, Google also announced that Gemini would also augment Performance Max across Google’s platforms, including email, search, and YouTube. The tech giant revealed that advertisers can create more engaging ads with extended headlines of up to 90 characters, a stark improvement from the previous limit of 30 characters.
The Controversies Behind AI Image Generators
In a Washington Post investigation conducted last year, prompts like “a productive person” resulted in pictures of fully white and almost entirely male figures, while a prompt for “a person at social services’ uniformly produced what looked like people of colour. As image generators are trained on large corpora of pictures and written captions to generate the best fit for a given prompt, they’re often prone to amplifying stereotypes.
Additionally, the report found that it’s a continuation of trends that have appeared in search engines and other software systems. For example, Jack Krawczyk, a senior director of products for Gemini at Google, stated that the company’s image generation capabilities reflect the tech giant’s “global user base,” and it takes “representation and bias seriously.”
Thus, Krawczyk followed up in a post on X: “We will continue to do this for open-ended prompts (images of a person walking a dog are universal). Historical contexts have more nuance to them. We will further tune to accommodate that.”
To catch the latest news of image generators such as OpenAI’s Sora, follow our Twitter page for daily updates and coverages of the trending stories filling your feed.