Mehul DasApril 14, 2022 16:07:16 IST
Think an AI system can’t replace human artists? Well, if you ask people in OpenAI, the developers of DALL-E, they will want to say something different.
Picture it – you have a great vision and can imagine the most beautiful pictures. You pick up a pencil or a painting brush but then suddenly realize that your drawing and painting skills are poorer than those of a kindergarten kid. A little embarrassing, isn’t it? Well, it has happened to all of us.
If someone tells you that now, you can just “draw” using your words, and how would anyone describe a painting? Well, DALL-E, an artificial intelligence (AI) system has achieved just that.
DALL-E is a new neural network AI system that can create kitsky photos and paintings, using only text prompts that the user can feed. Not only that, it can transform a finished painting and create countless repetitions based on the text prompt.
Developed by OpenAI, DALL-E was first introduced in 2021. The name is derived from Salvador Dali, a portmanteau of Pixar’s Wall-e and Dali, a surrealist painter. OpenAI’s developers need to fine-tune their machine learning algorithms and come up with a system that can create industry-leading works in detail and resolution.
OpenAI is not yet selling its system and is only accepting requests on their social media DM. The images they create are also not available for sale, but users can download images created for personal use. They are not making any pictures of real people to save themselves from possible lawsuits.
That being said, some of the things that the AI system has created can fool even the biggest fans of the industry into thinking that what they are seeing is the work of a skilled painter.
Now, like all things machine learning involved, there are some issues with DALL-E. First, it relies on properly tagged images on the Internet, which was a nightmare for people on OpenAI.
Second, as many Instagram users have noted, the ability to digitally create false images of realistic faces opens a Pandora’s box in the case of fake news.
And then there are the machine learning models that use OpenAI. Because their system relies heavily on images found on the Internet, which often contain some rather vitriolic material, the images that are rendered are often sexist and racist.
To address this, OpenAI requires that the use of their images include expressing their status as an AI-generated image. They add a color bar to the bottom right corner of the images.