This AI Program Can Draw Impressive Images Of Whatever You Tell It To

The DALL-E 2 program can churn out both realistic and creative images from a mere text description. But the OpenAI lab is well aware the technology could be easily abused too.

By Michael Kan

A new AI program is wowing the public with its ability to draw realistic and creative pictures from a mere text description.

The program is called DALL-E 2, and it comes from the San Francisco–based OpenAI research lab. On Wednesday, the lab introduced the AI by showing off the art it’s able to generate.

The results are impressive, and make us wonder if this is the future of image editing. For example, if you tell the AI to draw “a koala dunking a basketball,” DALL-E 2 will do just that by creating a photorealistic image of the animal flying through the air to a basketball hoop.

The AI can also edit existing photos by replacing the imagery inside. For example, if you type in “cute cat” for a dog photo, DALL-E 2 will transform the canine into a kitty.

In addition, the program can look at an original image and create different variations of it in the same style.

OpenAI created the DALL-E 2 by programming the AI to study existing images, which have been labeled with a text description explaining the person, animal or items depicted. Through “ deep learning “ training, DALL-E 2 can not only recognize what it’s inside an image, but also attempt to replicate it.

OpenAI introduced the first version of DALL-E back in January 2021. However, the second iteration has been upgraded to draw the pictures more accurately and with higher-resolution art quality.

It’s also important to note DALL-E 2 can draw pictures of things that don’t technically exist in the real world. For example, if you tell it to draw “monkey paying taxes,” the AI is smart enough to generate the animal facing a computer, even though it’s never encountered such an image during the training process.

The CEO of OpenAI Sam Altman said the DALL-E 2 represents a preview of what future artificial intelligence will have in store for computing.

“This is another example of what I think is going to be a new computer interface trend: you say what you want in natural language or with contextual clues, and the computer does it,” he wrote in a blog post. “We can imagine an ‘AI office worker’ that takes requests in natural language like a human does.”

DALL-E 2 also shows artificial intelligence is capable of creative work, not just automation. However, Altman said: “Although the upsides are great, the model is powerful enough that it’s easy to imagine the downsides.”

Altman didn’t go into the downsides. But past research and demos have shown that AI-created deep fakes of people could be exploited to spread propaganda and mislead the public.

Altman added: “Hopefully this summer, we’ll do a product launch and people will be able to use it for all sorts of things. We wanted to start with a research launch to figure out how to minimize the downsides in collaboration with a larger group of researchers and artists, and to give people some time to adapt to the change.”

OpenAI has published a research paper with more information on how DALL-E 2 was developed. It’s also created an Instagram account that features art DALL-E 2 has been creating.

Originally published at https://www.pcmag.com.