Google has introduced a series of new features to its Gemini AI, including a personalization tool called Gems, which allows users to customize the AI chatbot for specific tasks. This new feature enables users to tailor the Gemini chatbot to their needs, whether as a workout partner, a coding assistant, or a writing companion.
To create a personalized Gem, users can provide instructions on the desired style of responses, save a custom introduction, and even assign a specific character to the chatbot. Once these preferences are set, the customized Gem is activated and ready for use. This feature will be available exclusively to Gemini Advanced subscribers.
In addition to the customizable Gems, Google is also launching several predesigned Gems for broader tasks such as troubleshooting code, offering writing tips, and explaining complex topics in simpler terms.
Google is also rolling out the next-generation image generation tool, Imagen 3. This update includes the reactivation of Gemini’s ability to generate AI images of people—a feature that was previously disabled due to the creation of historically inaccurate images. The company has now implemented safeguards to prevent such issues in the future. These guardrails are designed to avoid overcorrection for diversity, which previously led to embarrassing mistakes.
“We don’t support the generation of photorealistic, identifiable individuals, depictions of minors, or excessively gory, violent, or sexual scenes,” stated Gemini Product Manager Dave Citron. He acknowledged that not every image generated by Gemini will be perfect but emphasized the company’s commitment to continuous improvement based on user feedback.
Additionally, Google has incorporated the SynthID tool to watermark images created by Imagen 3, ensuring the authenticity and traceability of AI-generated content.
Imagen 3 will be available to all users starting this week, though the ability to generate images of people will initially be limited to paid subscribers.