Google does a lot of insane things with its powerful artificial intelligence, and it has no intention of slowing down. The company just introduced us to an image generator. It’s called “Imagen”, and it allows Google to create photorealistic images with only text.
Senior Fellow & SVP at Google AI Jeff Dean posted a tweet that looks like a typical photoshopped image of a panda rolling dough. However, that image was completely generated using Google’s AI. This means that the panda in the image does not exist. Even photoshopped images need an existing subject. Looking at the image, it’s extremely impressive that it was created from AI.
AI can unlock joint human/computer creativity! Imagen is one direction we are pursuing:https://t.co/LTlE3pqq4WAdvertisement
"A high contrast portrait of a very happy fuzzy panda dressed as a chef in a high end kitchen making dough. There is a painting of flowers on the wall behind him." pic.twitter.com/SrqEv9jeHf
— Jeff Dean (@🏡) (@JeffDean) May 24, 2022
The official page for the technology explains a lot about what Google wants to accomplish with Imagen. It also has the research paper and other resources to help you gain more context about Imagen. However, if you visit the page, you’re most likely doing so to see examples of the technology (and there are a lot of them).
Some of the images include a small dog inside a house made of Sushi, a Google logo over Toronto, a cactus wearing sunglasses and a hat, two robots drinking with the Eiffel tower in the background, and the list goes on. Each of the images is extremely convincing. You’d have to look at details like shadows, reflections, and other small details to see the flaws in the images.
But, you can’t use it just yet
The ability to basically type words and get a faithful visual representation is a huge double-edged sword. Yes, it’d be great for having fun. Who wouldn’t want to type in “Avocado wearing a dress in space with a guitar”? You could probably have hours of fun just letting your “imagen-ation” flow free and coming up with all sorts of images.
The issue is that being able to “type in” images puts a lot of power in the wrong hands. While you’re creating images of avocados in dresses, people could type in any sort of inappropriate images involving children.
If you want to have a picture of a popular public figure for anything, you have to pay them. However, if you could just type in an image, you’ll be able to create and use images of those figures without paying them.
Also, if you could just create images, artists and graphic designers will lose business by the boatload. It could be the downfall for tons of artists who rely on the business. It’s for the best that Google keeps Imagen as a dream, as the world just isn’t ready for it.