DALL-E 2, How an AI is going to replace all artists?

DALL.E 2 is a highly sophisticated deep learning software that draws inspiration from a massive structured data source. OpenAI announced their program as an Artificial Intelligence (AI) capable of “creating realistic and artistic visuals from natural descriptions.” This last one allows you to generate visuals from textual queries. Previously, this tool was only available to professionals who needed to be added to a waiting list. With the recent release of their beta version, a million internet users can now let their imagination run wild and become herb artists.

“More than 1.5 million users are now actively creating more than 2 million photos every day using DALL.E 2 from artists and creative directors of famous brands to authors, architects.”

OpenAI

How does it work?

DALL.E 2 allows users to create images based on their preferences and make real-world changes to existing images with a simple scripting request. In other words, it is unnecessary to know how to draw; the software allows everyone to be an artist as long as they can articulate their idea well.

Here an example of an image we created with DALL-E 2, this is a cat dressed as Napoleon holding cheese

Will Dall-E 2 outperform artists?

Pierre Stock, an AI researcher at Facebook, believes that the development of picture-generation AI should help artists rather than replace them.

“We believe that the emergence of this form of AI will democratize access to interesting generational models.(…) Every artist is inspired by what others have done before them. It’s the same with the machine; one can be inspired by its output.”

Pierre Stock, IA researcher at Facebook

As a result, DALL.E 2 would be an additional tool to facilitate picture generation and similarly to Photoshop or Canva, it may reveal a significant advantage in producing low-cost graphics.

What are the stakes and limits of such a technology?

First and foremost, there is a legal gap in authorship rights: the various parties involved in the creation process make attributing an artwork to a single person problematic. Finally, there is the problem of deepfakes, which are synthetic images that can contain incorrect information and hoaxes. According to the experts, one solution would be to do a serious job of outreach and education to explain the algorithm’s operation and limitations clearly.

Leave a Reply