The Dall-E Mini image generator’s ridiculousness might be its main appeal
Salvador Dali, the Spanish artist born in 1904, is known for his surrealist paintings: melting clocks, elephants with insect-thin legs, distorted human faces and limbs. Dall-E Mini (pronounced like Dali), a new AI program, debuted in June. It produces images through user-generated requests such as a bottle from Ranch testifying in court , wikihow instructions for how to eat a and a “Silent Hill” nurse eating a . pizza.
The images are entertaining, and the program is gaining traction online for its funny art. You can enter any phrase or phase and then see the AI-generated image that was created. Dall-E Mini is an open-source project based on the original Dall-E technology from OpenAI, an AI research laboratory, which generates realistic images and art from text. Many of the Dall-E Mini’s art has been met with laughter. However, as technology like this becomes more refined and widespread, the risk for misuse increases. This is not something to be laughed at.
Dall-E Mini followed Dall-E 2, which OpenAI officially released in April 2022. Dall-E 2 works by creating associations between billions upon billions of images online and their descriptions. Dall-E Mini works in a similar way, but on a smaller scale. The project, led by Boris Dayma, used about 15 million images from three data sets to inform their model, and is roughly 27 times smaller than OpenAI’s original Dall-E program. Dall-E Mini is hosted on Hugging Face, a company that provides machine-learning models and tools and says they’re on a mission to “democratize good machine learning.”
Users have the ability to create any image they like, but the results tend to be more comical than accurate. Former head of AI at F5 technology security company, Shuman Ghosemajumder says that the humor and appeal of AI is partly due to the unexpected imperfections. Ghosemajumder explains that the images aren’t perfect, which is part of what makes it so much fun. “There are impressionistic images and nightmarish images. There are also crazy images.
All of this is fascinating to look at and fun to share.
Ghosemajumder said he can see this type of technology moving through several phases under the public’s eye. People are initially curious about the potential of a program like this. There’s wonder and learning as people explore the complexity of what they can create–as has been shown already online. He says that the second phase is more about people understanding the value of what can be produced.
” You can create rudimentary illustrations to serve a purpose. This will allow you to understand the technology better and help you plan for the future.
Looking forward, he envisions that the third phase in how these technologies are perceived will be how they lay the foundations for even more advanced innovations, such as high-quality synthetic videos.
However, as these innovations improve, so does the risk of misleading and dangerous images.
At the moment, Ghosemajumder states that Dall-E Mini’s images have a low quality enough that users know they are AI-generated and not an actual photo of, for example, a bottle of ranch that was being used as evidence in court. It will become more difficult to identify if something was created on an AI platform or captured in the nondigital world. This will make it easier to create images for misinformation campaigns online.
” It won’t be an impressionistic version or Darth Vader that people create. “Instead it’s going to be like, ‘wow, this looks like a real person doing something.’ I could create an image of a politician doing something that they never did, and eventually create an entire storyline and use that to disseminate disinformation.”
At the moment, Dall-E 2 has not been made available to the public. OpenAI’s developers are still pondering the risks of misuse. (Dall-E Mini, a similar but open-source version of the program, is also available. Similarly, Imagen, Google’s text-to-image program, has not released access to their platform. Both companies have employed artists and researchers to test the programs, improve operations and train data, make the art better, and ensure safety. Ghosemajumder states that a limited release is better than a general release. This principle is widely accepted in software development. He believes this will become more important as technology becomes more complex.
In the end, one of the greatest strengths technologies like Dall-E Mini have is that they will allow more people to visualize things easier than they can now, Ghosemajumder states. It is the democratization and accessibility to high-quality content.
” It unlocks people’s creativity, and allows people to communicate better,” Ghosemajumder states. “This has the potential for people to be more efficient and productive in creating illustrations and photos of various concepts they want to visualize .”
As Ghosemajumder said, his Dall-E Mini creation showed Chewbacca going about his day if he had a job.