Dubbed as “DALL-E”, this AI-bot was recently showcased by the non-profit AI organization, OpenAI. The name, if you are thinking is familiar, is actually a blend of the name of the popular Spanish artist, Salvador Dali (repopularized by Money Heist) and the name of the adorable bot from Pixar’s universe, Wall-E. Now, DALL-E is not your regular text-to-image conversion AI. It is a unique neural network that takes your descriptive text sentences and turns them into accurate pictures. So, in a way, it is like giving descriptions of a scenario or an object to a sketch artist for him/her to create the perfect image that is in your mind.
How Does it Work?
Now, as the researchers state in their official blog post, “DALL·E is a 12-billion parameter version of GPT-3 trained to generate images from text descriptions, using a dataset of text-image pairs”. In their testing, the researchers found that the model has numerous unique capabilities to convert texts into surreal yet goofy images. If you check out their official blog post, you will find that there are some interactive examples provided by the researchers to show how the model works. These examples already have pre-set text sentences for the model to render pictures. However, there are some keywords in these sentences that you can change to make the bot render different pictures. These keywords mostly involve descriptive words that help DALL-E understand the color, shape, pattern, or texture of an object. So, as soon as you change these words from a drop-down list of other similar words, the pictures instantly change based on the selected word. Moreover, as per the creators, the model improves itself the more you use it.
Now, it is pretty impressive to see an AI understand the exact shapes, colors, and textures of objects to create accurate images. Now, I am just waiting for the day when these kinds of AI-models would be used as actual sketch artists in public organizations like the Police department.
