Google finally adds AI text-to-image generation, but it’s not where you think


Example of Google SGE interface


Since the generative AI craze began, Google has been trying to leverage its position as the leading search browser to spark interest from the public in its AI advances. Now, Google is adding more generative AI features to search. 

On Thursday, Google unveiled two new features for search: Text-to-image generation and a written drafts feature in its AI-powered Search Experience (SGE). 

Also: 80% of enterprises will have incorporated AI by 2026, according to a Gartner report

Although Google is not new to AI image generation, announcing its own Imagen text-to-image model as early as March 2022, it has yet to release an image-generator to the public — until now. 

Users that have opted in SGE will now be able to create images directly in Google. All the user will have to do is enter the description for the image they’d like generated in the search bar. Then, within seconds, Google will generate four image options.

The demo shows a user entering the prompt “draw a picture of a capybara wearing a chef’s hat and cooking breakfast” into the search bar and immediately being shown four images. 


Once a user clicks on one of the four images, Google will show an expanded initial query that contains descriptive details that users can tweak further to get their ideal result. This will help the prompt writing process, which can be tricky when using AI models. 

In this example, one of the images had an expanded query that read, “a photorealistic image of a capybara wearing a chef’s hat and cooking breakfast in a forest, grilling bacon.” If a user wanted to change the background, they could quickly edit that one word in the expanded query.

Also: The impact of artificial intelligence on software development? Still unclear

SGE users might also see an option to create AI-generated images directly in Google Images, which could be useful when searching for something in particular and not being able to find it in your Images results, especially ideas or creative requests.  

Google reassures users that they are rolling out this feature in a responsible way, including safeguards that prevent the generation of harmful or misleading content and metadata labeling and watermarking that indicates it was generated by AI. 

The company will also be releasing a tool called “About this image,” which is meant to help people discern the context and credibility of an image by showing them where a similar version was first seen by Google or other pages that use a similar image. 

The fact that Google is implementing a much anticipated AI text-to-image generator to Search is puzzling because Google Bard, its AI chatbot, has yet to showcase text-to-image generative capabilities despite competitors like Bing Chat that do. 

Also: The best AI image generators: DALL-E 2 and alternatives

It perhaps signals that Google is less interested in developing Bard and making it a leading AI chatbot and would prefer to build out its search engine to be infused with AI throughout. 

Google is also adding “written drafts in SGE,” which essentially adds a text generator to Search to help you with drafting written content like emails or quick notes. From the demo image, it looks like users will be able to select the length and tone of the text they want. 

This feature is rolling out to SGE experiment users in English in the US. If you haven’t signed up, all you have to do is visit the Search Labs page and log into your personal Google account. 


Source link