Google is rolling out early access to its latest text-to-image AI, Imagen 3, to more people. If you’re a Gemini Advanced subscriber, you might see a popup in the beta version of the Gemini and Google apps inviting you to try it out. Previously, only a handful of creators had access, but now Google is letting more users take the AI for a spin. It’s still in the testing phase, so not everyone will get it and you might need to wait longer for a public release.
Imagen 3 is the best AI image generator yet
Imagen 3 is Google’s top-tier text-to-image AI model. The company first announced it on May 14, and it was designed to create stunning, lifelike images from text descriptions — and it’s doing just that. Previews of images generated with the AI during its announcement were vivid. You couldn’t even tell they were created with AI.
Google released samples of what Imagen 3 can do, with prompts like “A pair of well-worn hiking boots, caked in mud” and “Three women stand together laughing”. There’s also “A close up of a sleek wolf perched regally in front of gray background”, which you’d mistake for a real wolf shot with a camera if you weren’t told the truth of the AI behind it.
Select Beta testers of the Google and Gemini apps should see a “First Look” notification prompting you to take advantage of Imagen 3. Also in the works is Veo, a highly advanced video generation model. Veo is capable of creating high-definition 1080p videos in various cinematic styles. It can also produce consistent and coherent footage based on your prompts, understanding cinematic terms and rendering realistic movements. The AI is currently available to select creators through a private preview.
Google’s Imagen 3 raises ethical concerns
AI-generated content often lacks depth and meaning. At first, they were impressive. The thought of creating a unique image with a random combination of words intrigued everybody. Then we got bored with the hollow eyes and distorted fingers pumped out to the internet. It made them unmemorable and disposable. Google hopes to fix that with Imagen 3 with highly realistic details, and more convincing deep fakes.
Imagen 3 is our highest quality text-to-image model. It generates an incredible level of detail, producing photorealistic, lifelike images, with far fewer distracting visual artifacts than our prior models. Google hopes to fix that with Imagen 3 with highly realistic details, and more convincing deep fakes.
Interesting as it sounds, it does bring concerns when you consider what “photorealistic” and “convincing deep fakes” means for the future. You’ll see more images promoting false news or propaganda, non-consensual explicit content as tools for manipulation, and let’s not forget the argument of ownership and attribution with AI-generated images based on existing works.
Of course, no one is pointing a finger at Google to say they’ve created a dangerous tool. But Robert A. Heinlein did say that “There are no dangerous weapons; there are only dangerous men”. The danger doesn’t lie in the technology itself, but in the intentions of those who use it.
