DALL-E-2: The Future of Art?

Text prompt: “A painting in the style of Monet”

Jiajun Shen

DALL-E-2, a machine learning model that generates digital images, is a recent innovation from the creative minds at OpenAI, an artificial intelligence research company. They have created what is essentially a text to picture AI, which generates images based on text prompts. The algorithm creates high quality images so impressive, that the AI art has been been compared to the art of professional artists. Besides photorealistic images, DALL- E-2 can also recreate art styles such as watercolor and oil paintings. DALL-E-2 is quite impressive, but how does it measure up to real artists? In a recent art contest at the Colorado State Fair, Jason Allen used AI generated technology to create digital art, winning the top prize of $300.

Text-to-image bots, such as stable diffusion and DALL-E-2’s predecessor, DALL-E, was first introduced last year. DALL-E-2 uses a modified version of GPT-3, which is an autoregressive transformer language model algorithm also created by OpenAI. This technology can edit images and display how how elements behave and should be placed in the real world. It can do all of this very quickly, having a response time of 10 seconds. Still, today’s AI generated art is far from perfect, struggling at times with generating text and sentences into pictures, turning the words into gibberish.