Google on Wednesday announced the release of customizable “Gems” and the first major update to its flagship AI model, Gemini, in an effort to challenge OpenAI's dominance in the AI market.
Gemini also directly integrates the company's AI image generation tool, Imagen 3. However, it doesn't generate images of people, as the previous release produced problematic images that forced Google to take the tool offline.
First announced by Google at Google I/O in May, Gems allows users to create a set of modules for customized AI assistants built on the Gemini model that can be used for a variety of projects, from coding to career advice, similar to the GPT feature in OpenAI's ChatGPT. According to Google, Gems is available to Gemini Advanced, Business, and Enterprise users.
“With Gems, you can create a team of experts who can help you think of a tough project, brainstorm ideas for an upcoming event, or write the perfect caption for a social media post,” Google said in a statement. “Gems can also remember detailed instructions, saving you time on tedious, repetitive, or difficult tasks.”
For creators who can't think of the exact phrase to create their Gems, Gemini also features an AI-powered rewrite feature that allows you to fine-tune the prompts for creating your Gems. The output of Gems can be shared via social media links or shared on Google Docs or Gmail to add to email drafts.
“When it comes to sharing, currently any gems you create are for personal use only,” a Google representative told Decrypt. “You can share chats with your gems by creating a public link, but others you share the link with will not be able to continue chatting with your gems.”
This is a more limited offering than OpenAI's GPT, and can be shared more fully with others who can use the same customizations.
The integration of the Imagen 3 image generator also expands Gemini's built-in capabilities, with the tech giant reiterating its cautious approach to deployment.
“We conduct extensive red team testing internally and externally and engage with independent experts to ensure continuous improvement,” a Google representative said. “We have prohibited uses policies and prohibit responses that violate our policies.”
Google released Imagen 3 earlier this month after first announcing it in May, and the product faces stiff competition from tools like OpenAI, Midjourney and Flux's Dall-E, which is built into Elon Musk's Grok chatbot.
Image created by Decrypt using AI
“Imagen 3 sets a new standard for image quality by generating images with just a few words,” Google said. “You can also ask Gemini to create images in a variety of styles, including photorealistic landscapes, textured oil paintings, and whimsical claymation scenes.”
Gemini can draw pictures of animals and objects, but they are not yet able to draw pictures of people.
“People image generation is coming soon to Gemini Advanced,” the chatbot responds upon request.
“With Imagen 3, we've made significant efforts to improve the user experience when generating images of people,” Google said. “We no longer support generating graphic, personally identifiable images, depictions of minors, or excessively cruel, violent, or sexual scenes.”
“Of course, like any generative AI tool, not all images Gemini creates are perfect, but we're listening to feedback from our early users and will continue to improve it,” Google added. “We'll be rolling this out gradually, with the goal of making it available to more users and languages soon.”
Generally intelligent newsletter
A weekly AI journey narrated by generative AI model Gen.
