Google’s image generation feature for its AI tool Gemini, which was part of the larger suite formerly known as Bard and Duet, came under fire earlier this year in February 2024.
Here’s a deeper dive into what happened:
- Inaccurate and Offensive Images: Users discovered numerous inaccuracies in the images created using Google Gemini, particularly in depicting historical figures. These inaccuracies appear to be an attempt to address longstanding issues of racial bias in AI systems. Despite efforts to mitigate bias, Google’s Gemini tool has swung too far in the opposite direction, producing historically inaccurate images of women and people of color. For instance, prompts to depict America’s founding fathers resulted in images showing a woman of color signing the United States Constitution. Similarly, images shared on social media depicted people of color portrayed as Vikings, Nazi soldiers, and even the Pope, despite the prompts not requesting such alterations.
- Historical Depictions & Nuance: There were additional concerns beyond ethnicity. Gemini struggled with sensitive historical events. Users reported it wouldn’t generate images for prompts like the Tiananmen Square massacre or the Hong Kong protests, citing the inability to capture the gravity of the situations. This raised questions about the tool’s ability to handle complex topics.
- User Outrage & Media Attention: The inaccurate image generation sparked outrage on social media platforms. News outlets like Al Jazeera and Fox Business covered the story, highlighting the tool’s biases and the offensive nature of some outputs.
Following the controversy, Google took these steps:
- Pausing the Feature: In February 2024, Google made the decision to temporarily halt the image generation feature of Gemini. This allowed them to focus on fixing the underlying problems.
- Explanation and Apology: Google released a blog post acknowledging the issues and apologizing for the experience users had. They explained how the tool was designed to avoid specific biases, but those efforts backfired in unintended ways.
- Working on Improvements: Google is currently working on improvements to the image generation feature. Their aim is to ensure it generates accurate, unbiased images and can handle sensitive topics with more nuance.
First, our tuning to ensure that Gemini showed a range of people failed to account for cases that should clearly not show a range. And second, over time, the model became way more cautious than we intended and refused to answer certain prompts entirely — wrongly interpreting some very anodyne prompts as sensitive.
These two things led the model to overcompensate in some cases, and be over-conservative in others, leading to images that were embarrassing and wrong.
“Gemini image generation got it wrong. We’ll do better.”
While there’s no official relaunch date yet, Google has indicated they’re working diligently to bring the feature back with better safeguards in place.
What about Meta’s AI image generator, Imagine?
In March 2024, Imagine faced similar issues to Google’s Gemini. There were reports of historically inaccurate outputs. For example, prompts for “a group of popes” might generate images of Black popes, even though there haven’t been any Black popes in history. This sparked concerns about potential biases in the training data.
Related: Where can you download free stock photos for your website?
Where can you download free stock videos for your website?
Where to get photos/images for your website?
Should you avoid using stock photos on your website?
Are you ready to design & build your own website? Learn more about UltimateWB! We also offer web design packages if you would like your website designed and built for you.
Got a techy/website question? Whether it’s about UltimateWB or another website builder, web hosting, or other aspects of websites, just send in your question in the “Ask David!” form. We will email you when the answer is posted on the UltimateWB “Ask David!” section.
Copyright ©