Since January 2021, advances in AI research have produced a large number of deep learning models capable of generating original images from simple text cues, effectively expanding the human imagination. Researchers at OpenAI, Google, Facebook and others have developed text-to-image tools that have not yet been released to the public, and similar models have proliferated online in the open source field and in smaller companies such as Midjourney.
These tools represent a massive cultural shift because they eliminate the need for technical labor from the image creation process. Instead, they select the creative idea, the skillful use of language, and the curatorial taste. The latter consequences are difficult to predict, but as the invention of the camera and the digital camera later, these algorithms herald a new form of democratized expression that will start another explosion in the volume of human-produced images. But like other automated systems made up of historical data and images from the Internet, they also present unresolved risks.
The video above is an introduction to how we got here, how this technology works, and some of the implications. And for an extensive discussion of what this means for human artists, designers, and illustrators, watch this additional video:
You can find this video and everything Vox videos on our YouTube channel.
