Google apologizes as AI diversity feature backfires
CALIFORNIA, UNITED STATES — Tech giant Google has apologized for problems with the rollout of new artificial intelligence (AI) image generation technology in its Gemini chatbot.
The company acknowledged the mistake, adding it had tried to make the tool showcase diversity, but this led to inappropriate or inaccurate images in some cases, making them “pause the image generation of people and will re-release an improved version soon.”
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. https://t.co/SLxYPGoqOZ
— Google Communications (@Google_Comms) February 22, 2024
The issue came to light when users on social media shared examples of the AI generating images that depicted minorities in historical settings where they would not have been present.
For instance, it showed a Black woman as a founding father and Asian and Black people as Nazi-era German soldiers.
Google introduced the image feature about three weeks ago, building it using an earlier AI system called Imagen 2. In a 2022 paper, Imagen’s creators had warned it could be misused to spread misinformation or cause harm.
In a blog post, Google executive Prabhakar Raghavan admitted the image tool “missed the mark.”
“Some of the images generated are inaccurate or even offensive. We’re grateful for users’ feedback and are sorry the feature didn’t work well.”
Our AI image generation feature got it wrong. Here's more on what happened and how we're working to fix it. https://t.co/CUrCHimu1s
— Google Public Policy (@googlepubpolicy) February 23, 2024
He said when tuning the software, the company “failed to account for cases that should clearly not show a range” of diversity. Also, the system started refusing reasonable prompts, wrongly thinking they were 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.”
Raghavan said Google will thoroughly test the feature before allowing images of people again. Critics, including tech billionaire Elon Musk, argue the issues show current AI still has serious limitations despite rapid progress.
I’m glad that Google overplayed their hand with their AI image generation, as it made their insane racist, anti-civilizational programming clear to all
— Elon Musk (@elonmusk) February 23, 2024
