Google is re-enabling its image generator feature for some Gemini AI users after the tool was temporarily suspended due to concerns that it could not reliably represent white people.
The Alphabet-owned technology company said in a blog post on Wednesday that the latest generation of its text-to-image tool, Imagen 3, will soon be available to users who pay for subscriptions to Gemini Advanced, Gemini Business and Gemini Enterprise.
“In the next few days, we will also start generating people images, with an early access release for our Gemini Advanced, Business and Enterprise users, initially in English. We have been working to make technical improvements to the product, as well as improved evaluation sets, red teaming exercises and clear product principles,” the company said.
Users of the tool can write text prompts that Imagen 3 converts into visual images. In its blog post on Wednesday, Google provided a sample prompt and a corresponding image.
“An animated image of a tiny dragon hatching from an egg in a sunlit meadow, surrounded by curiously glowing butterflies. Bold colors, detailed scales,” was the prompt for the science fiction theme.
In another example, Google instructed the tool to generate an image of a “ball gown made from paper napkins in an elegant showroom.”
Google added that Imagen 3 has new built-in safeguards to ensure it complies with the company’s design principles, which include that Gemini cannot create “instructions for self-harm” or, in the case of Imagen 3, “pornography or excessively gory images.”
Google came under fire when it first introduced its text-to-image tool and prompts for users to generate historical figures such as German soldiers in World War II and popes, all of whom were historically white and male. Some of Gemini’s images depicted Nazi soldiers as Black and Asian and popes as female.
At that time, CEO Sundar Pichai named its AI app problems “unacceptable.”
Google assured users that this time it had made “significant progress in providing a better user experience when generating people images.”
“We do not support the generation of photorealistic, identifiable people, the depiction of minors, or excessively gory, violent or sexual scenes,” the company said in a statement. However, there was a caveat, as the AI-based tool is still novel and under development.
“Of course, as with any generative AI tool, not every image created by Gemini will be perfect, but we will continue to listen to feedback from early users as we continue to improve. We will roll this out gradually and look to make it available to more users and languages soon.”
On Thursday, the image generation feature was not available to all Gemini Advanced subscribers.
When asked to depict a human, the tool responded: “Image generation of people is coming soon to Gemini Advanced.”
