What you need to know
- Google is making upgrades to Bard, and users can access the new features starting today.
- The company is bringing Imagen 2, its next-generation image generation model, to Bard and a new tool called ImageFX.
- Additionally, Google is expanding the reach of Gemini Pro in Bard, which is now available in more than 40 languages and over 230 countries and territories.
Google Bard, the company's AI chatbot, is getting smarter starting today thanks to the addition of Imagen 2 and Gemini Pro. These two artificial intelligence models will help Bard give better responses and higher-quality images.
Google announced Gemini Pro last year, and the new AI model has already appeared in Bard in some regions. However, Gemini Pro is now powering Bard worldwide in a significant expansion. In total, Bard is now available in more than 40 languages and over 230 countries and territories.
One of Bard's best features is double check, and it allows users to verify Bard's responses are accurate with one tap of the Google logo. This tool cross-checks Bard's answer against similar internet results to check for accuracy. It's designed to improve user confidence in Bard, since like all AI chatbots, it has the potential to hallucinate. Starting today, double check is available in over 40 languages, while it was previously restricted to English.
However, the bigger addition is Imagen 2, the text-to-image model Google revealed late last year. Here's what Google has to say about why Imagen 2 is an upgrade over the previous image generation model:
"Imagen 2 has been trained on higher-quality, image-description pairings and generates more detailed images that are beer aligned with the semantics of people’s language prompts," Google said in a press release. "It’s more accurate than our previous system at processing details, and it’s more capable at capturing nuance — delivering more photorealistic images across a range of styles and use cases."
Imagen 2 uses a diffusion-based model, and aims to provide better images than comparable image generation tools. For example, some image generators struggle with things like text, faces, and hands. Google says that Imagen 2 is improved specifically in creating faces and hands that are more realistic in nature.
For end users, the inclusion of Imagen 2 in Bard means that images generated from text prompts will be better-looking and will more accurately represent the initial prompt.
Google emphasized that Imagen 2 and Bard both use "responsible AI," which has come to mean many things. Imagen 2 will not create images that are violent, offensive, or contain sexually explicit content. Additionally, Imagen 2 has some training that will help it "avoid" creating images of specific people. However, based on that wording, it's possible users can create images of named people in some instances.
Any image generated with Imagen 2 will have a digital SynthID watermark. These aren't visible to users when viewing the AI images, but are still there and detectable.
Aside from appearing in Bard, the Imagen 2 model will generate images in Google Ads, Duet AI in Google Workspace, and the Search Generative Experience. Imagen 2 also powers ImageFX, an image generation tool currently available in Labs.
Today's upgrades to Google Bard fulfill a lot of promises made by Google last year. Next up, we'll be waiting for Bard Advanced to debut with the promised Gemini Ultra model. On the mobile side, there's also Assistant with Bard — or whatever Google ends up calling it.
