Mehul DasApr 14, 2022 4:07:16 PM IST
Think an AI system can’t replace human artists? Well, if you ask the folks at OpenAI, the developers of DALL-E, they wouldn’t agree.
Imagine this – you have wonderful vision and can imagine the most picturesque paintings. You pick up a pencil or paintbrush, but suddenly realize that your drawing and painting skills are inferior to those of a kindergarten child. A little embarrassing, isn’t it? Well, it’s happened to all of us.
What if someone told you that now you can “paint” just by using your words, and how would a painting be described? Well, DALL-E, an artificial intelligence (AI) system has achieved this goal.
DALL-E is a new neural network AI system that can generate kitschy photos and paintings, just using text prompts a user might give it. Not only that, it can modify a finished painting and generate many iterations, again, based on text prompts.
Developed by OpenAI, DALL-E was first introduced in 2021. The name is a portmanteau of Pixar’s WALL-E & Dali, by surrealist painter Salvador Dali. It took OpenAI developers to refine their machine learning algorithms and come up with a system capable of creating original works of art with artful detail and resolution.
OpenAI does not sell their system yet and only accepts requests through their social media DMs. The images they generate are also not available for sale, but users can download the images they generate for personal use. They also do not generate images of real people, in order to protect themselves from possible lawsuits.
That being said, some of the things generated by the AI system would fool even the most serious art connoisseurs into believing that what they are seeing is the work of a master painter.
Now, as with anything involving machine learning, there are also some issues with DALL-E. First, it relies on images on the internet being properly tagged, which was a nightmare for the OpenAI folks to manage.
Second, as several Instagram users have pointed out, the ability to digitally create realistic fake images opens Pandora’s Box of fake news.
And then there’s the machine learning model that OpenAI used. Because their system relies heavily on images from the internet, which often contain rather vitriolic elements, the rendered images are sometimes rather sexist and racist.
In order to address this issue, OpenAI requires that use of their images include disclosure of their status as an AI-generated image. They also add a color bar in the lower right corner of the images.