Deep fake Taylor Swift, Google releases a new AI-powered tool, ImageFX, to create images.
Powered by Imagen 2, a GenAI image model developed by Google’s DeepMind team, ImageFX offers a prompt-based user interface for creating and editing images. This is no different from tools like OpenAI’s DALL-E 3, Midjourney, Meta’s Imagine with Meta AI, and Microsoft Designer. But ImageFX’s unique twist is its “expression chips” — basically a list of keyword suggestions that allow users to experiment with “adjacent dimensions” of their creations and ideas.
“Designed for experimentation and creativity, ImageFX lets you create images with a simple text prompt, then easily modify them with a new approach to prompts using expression chips,” Google writes in a suspension.
But what about the potential for abuse — especially in light of recent events?
Image Credits: Google
Google claims it has taken steps to ensure ImageFX cannot be used in ways it was not intended, for example by adding “technical safeguards” to limit “problematic results” such as violent, offensive and obscene sexual content. ImageFX also has a prompt-level filter for “famous people,” presumably public figures — though Google wasn’t particularly clear on that point in its press release.
“We’ve invested in the security of education data from the ground up,” Google said. “In line with our AI principles, we have also conducted extensive adversarial testing and red teaming to identify and mitigate potentially harmful and problematic content.”
As an added security measure, Google tags images produced using ImageFX with SynthID, a digital watermark that is said to be resistant to image manipulation and cropping.
A sample image from Imagen 2. Image Credits: Google
“SynthID watermarks are imperceptible to the human eye but detectable for identification,” Google continues in the blog post. “With additional information in ‘About this image’, you’ll know if an image may have been created with Google’s AI tools when you come across it in Google Search or Chrome.”
You’ll find ImageFX in the AI Test Kitchen, Google’s web app for experimental AI projects.
Image 2 expanded
In related news today, Google said it’s bringing Imagen 2 to more of its products and services starting this week, including its next-generation AI search experience and Vertex AI family of managed AI services.
Imagen 2 — which also now provides text-to-image capabilities in Google Ads and Duet AI in Workspace, Google’s suite of GenAI products for productivity — has entered Google’s Search Generative Experience (SGE). SGE, which started showing image creation tools for users on Google Image Search last October, is now tapping Imagen 2 to create images. Users can enter a prompt specifying what kind of image they want and SGE will display four results directly in the SGE chat experience.
Another sample from Imagen 2. Image Credits: Google
In Vertex AI, Imagen 2 is available through an API to Google Cloud customers. Elsewhere, Imagen 2 is now invoked through Bard, Google’s AI-based chatbot.
“With Imagen 2, Bard understands simple or complex messages so you can create a series of high-quality images,” explains Google. “Just type in a description—like ‘create an image of a dog riding a surfboard’—and Bard will create custom, wide-ranging graphics to help bring your idea to life.”
Google has yet to reveal the data it used to train Imagen 2, which – while disappointing – is not surprising. It’s an open legal question whether GenAI vendors like Google can train a model on publicly available—even copyrighted—data and then reverse engineer and commercialize that model.
Image Credits: Google
Related lawsuits are playing out in the courts, with sellers arguing that they are protected by the fair use doctrine. But it will be a while before the dust settles.
Meanwhile, Google is playing it safe by keeping quiet on the matter.
