Photo by Lee Unkrich, one of the famous Pixar artists, as a 7th grader. He is looking at a picture of a train locomotive on the screen of the first computer at his school. Oh, he thinks. Some of the magic ends, however, when Lee learns that the picture didn’t appear by simply asking for a “train picture”. Instead, it needed to be carefully written and delivered—by hardworking people.
Now picture Lee 43 years later, stumbling upon DALL-E, an artificial intelligence that creates original drawings based on ideas submitted by people that can be as simple as “a picture of a train.” As he writes in the text to create picture after picture, and Oh he is back. Only this time, it won’t go away. “It sounds like a miracle,” he said he says. “When the results were seen, my breath was taken away and tears started to well up in my eyes. That’s magic.”
Our machine has crossed the border. All our lives, we have been assured that computers cannot create reality. However, all of a sudden, millions of people are now using a new type of AI to create stunning, never-before-seen images. Most of these users are not, like Lee Unkrich, artists, and that’s the point: they shouldn’t be. Not everyone can write, direct, and edit like an Oscar winner Play Story 3 or That’s rightbut all he can activate the AI image generator and write an idea. What appears on the screen is amazing in its reality and depth. So the Universal answer: Oh. On just four services—Midjourney, Stable Diffusion, Artbreeder, and DALL-E—humans and AIs are now producing more than 20 million images per day. With a paintbrush in hand, artificial intelligence has become a wow engine.
Because these amazing AIs have learned their skills from billions of human-generated images, their output revolves around how we expect images to look. But because it is a strange AI, incomprehensible even to its creators, it recreates new images in a way that no human can imagine, filling in details most of us would not have the ability to think, let alone the ability. to do. They can also be instructed to create many variations of something we like, in any color we want – in seconds. This, in the end, is their strongest advantage: They can create new things that are coherent and logical but, at the same time, unexpected.
Unexpected and new images created by AI, in fact, that – surprisingly quietly after that Oh-Another thought is common to almost everyone who has experienced it: Man-made art must end now. Who can compete with the speed, affordability, scale, and, of course, sheer genius of this machine? Is creativity just another human trait that we must surrender to robots? And the next obvious question: If computers can design, what else can they do that we’ve been told they can’t?
I’ve spent the last six months using AI to create a lot of amazing images, often staying up all night trying to find them. only one the beauty hidden in the code. And after talking to developers, power users, and other early adopters of these generators, I can make it clear: Generative AI will change the way we make almost everything. Oh, and not a single human artist will lose their job because of this new technology.
No exaggeration to call images created with the help of AI cocreations. The worrisome secret of this new power is that its best use is not one-time typing but a long-term conversation between humans and machines. Each image’s progress comes from many iterations, back-and-forths, twists, and hours, sometimes days, of working together—all thanks to years of machine learning advancements.
AI image generators were born from the marriage of two types of technology. One was the history of deep learning neural networks that could create beautiful images, and the other was a natural language that could be used to connect a graphics engine. The two were combined into a language-driven image generator. Researchers cataloged all the images on the Internet that had words nearby, such as captions, and used billions of these samples to connect shapes to words, and words to colors. With this new integration, users can enter a number of words – quickly – that describe the image they want, and that prompt can generate an image based on that word.
