• AiNews.com
  • Posts
  • Google Restores AI-Generated People Feature in Gemini with Safeguards

Google Restores AI-Generated People Feature in Gemini with Safeguards

Author

Alicia Shapiro
August 29, 2024 • Estimated Reading Time: 3 minutes

An image representing Google's reintroduction of the AI-generated people feature in Gemini, powered by the upgraded Imagen 3. The central focus is a digital interface where a user generates diverse human faces using the Gemini AI, with options for customizing the images. The interface also highlights built-in safety features to prevent inaccuracies. The background includes subtle Google branding and elements suggesting AI-driven creativity, such as digital brushes and image editing tools. The color scheme incorporates Google's signature colors—blue, red, green, and yellow—emphasizing innovation and safety in AI technology

Image Source: ChatGPT-4o

Google Restores AI-Generated People Feature in Gemini with Safeguards

After pausing the feature earlier this year due to concerns over historically inaccurate images, Google has reintroduced the ability to generate images of people through its Gemini AI chatbot. The feature, which had been disabled following reports of inappropriate content, will be rolled out to Gemini Advanced, Business, and Enterprise users in English over the coming days.

Powered by Imagen 3: Google's Latest AI Image Generator

This reintroduction is powered by Imagen 3, the latest version of Google’s AI text-to-image generator. Imagen 3, quietly launched earlier this month through Google’s AI Test Kitchen, is now available on the Gemini platform across all languages. The tool allows users to generate a wide range of images, from photorealistic landscapes to detailed oil paintings, with just a few descriptive words.

Addressing Previous Concerns with Built-In Safeguards

Google had initially paused the image generation feature in February after users discovered that Gemini created historically inaccurate and inappropriate images, such as racially diverse Nazis. The upgraded Imagen 3 model now includes built-in safeguards to prevent such errors. According to Dave Citron, Google’s senior director of product management for Gemini, the model “performs favorably compared to other image generation models available” and has undergone significant technical improvements, enhanced evaluation sets, and rigorous red-teaming exercises to identify and address vulnerabilities.

Limitations and Ethical Considerations

Google’s Gemini will not allow the generation of photorealistic images of public figures, content involving minors, or any gory, violent, or sexual scenes. “Of course, not every image Gemini creates will be perfect, but we’ll continue to listen to feedback from early access Gemini Advanced users as we keep improving,” Citron added. Google also plans to expand this capability to more users and languages in the near future.

Not to Be Confused with Google’s Remagine Feature

It’s important to note that this feature is separate from Google’s new Remagine tool, which allows users to incorporate AI elements into photos taken with the company’s new Pixel 9 phones.