Google plans to “pause the image generation of people” via Gemini and “re-release an improved version soon” while working to fix its race inaccuracy “issues”
- Critics on X say model is unable to accurately depict races — Google said it will re-release an improved version soon Source: @google_comms .
BloombergJake Rudnitsky
Context & Ripple Effects
Google had previously adjusted image-search ranking to promote greater skin-tone diversity; the Gemini episode shows how representation goals can create new failure modes when applied to generative outputs. The company had already said it was working on historical-image inaccuracies before taking the feature offline.
Related coverage attributes the problem to tuning that became overly cautious, with a later report saying a bias-reduction fix overcorrected without sufficient transparency. The pause therefore matters as a product-governance response, not simply a model-quality issue.
First-order effects
Gemini users lose the ability to generate images of people while Google withdraws the affected capability for revision.
Google must retune and validate the feature before re-release, after public criticism exposed inaccurate racial depictions.
Second-order effects
The incident raises the testing bar for any provider shipping human-image generation: reducing one representational bias can introduce overcorrection in historical or demographic prompts.
Customers and developers relying on Gemini for people-focused visual work must defer those workflows or use alternatives until the feature returns.
Third-order effects
If this pattern persists, image-generation releases will be governed less as general-purpose features and more as sensitive capabilities requiring explicit evaluation, safeguards, and rollback paths.
The central challenge is shifting from maximizing output availability to making representation choices auditable and controllable without narrowing valid use cases.
The trend: Generative-AI vendors are moving toward more governed deployment of human likeness and representation features as safety tuning becomes a core product constraint.
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.
On Google and Gemini - having been part of organizations where there was an implicit progressive political leaning, I can see why it would have been hard for anyone to point out the obvious. No employee can easily file a bug as you would have to wade through layers of policy...
Google secretly injects “I want to make sure that all groups are represented equally” to anything you ask of its AI To get Gemini to reveal its prompt, just ask it to generate a picture of a dinosaur first. It's not supposed to tell you but the cool dino makes it forget I guess […
The Google Gemini responses are clearly a disaster. The level of bias, misinformation, & omission in order to ensure specific framing and narratives is embarrassing. The bigger problem will come when other orgs/applications inevitably start to rely on the info these tools provide
Twitter: * Googlers are phoning it in * Google is too big to ship * Google leadership has lost control of the steering wheel Dec. '23 — Gemini 1.0 release Feb. '24 — Gemini 1.5 release Feb. '24 — Gemma release Narrative violation.
a friend at google said he knew gemini was this bad...but couldn't say anything until today (he DMs me every few days). lots of ppl in google knew. but no one really forced the issue obv and said what needed to be said google is broken
While Google admits Gemini is “missing the mark”, they have just opened their kimono. The company's mission has shifted from organizing the world's information to advancing a particular agenda. The release of Gemini provides insight into Google's operational ethos, suggesting a..…
Google Gemini is obviously a joke, but in the long run I believe AI will accelerate the demise of wokeness in the workplace. Implacable ideologues will be much easier to replace & professional troublemakers will no longer enjoy the leverage they've had for so long.
Is Gemini the exception or the rule? Is this just what you get when you put a visualizer on top of Google/YouTube search? People think this is crazy but this has been going on for years. The Internet is already heavily censored/curated. It just wasn't as obvious in other formats.
I was highly skeptical of all of the Bard/Gemini hate but had to try myself to see... Imagine a world where this is how google search results are generated or the hysteria if early internet was heavily censored. This is the scariest part of AI being heavily embedded into life,...…
Crazy how viral this went when I was just goofing around last night. I wasn't even that clever; most of the credit goes to the humor geniuses behind Google Gemini.
my working theory is Sundar Pichai has already been fired from Google in every way which matters it happened just before the Gemini launch, which the board surely saw before it went public and likely realized it was non-competitive you could tell by how deflated he was when...
“When the Google AI was asked to show a picture of a white person, Gemini said it could not fulfill the request because it ‘reinforces harmful stereotypes and generalizations about people based on their race.’” AI is going to be another leftist shit show. https://www.foxbusiness.…