On June 6, Cover facean open source hosting company artificial intelligence project, saw traffic to an AI visualization tool called DALL-E Mini skyrocketed.
The ostensibly simple app, which generates nine images in response to any typed text prompt, was launched almost a year ago by an independent developer. But after some recent improvements and a few viral tweets, the ability to crudely sketch all forms of surrealism, fun, and even fantasy has suddenly become magic. Take a look at its manifestations of “Thanos is looking for his mother at Walmart,“”drunkards roam around Mordor,“”Darth Vader’s CCTV camera footage is groundbreaking,” and “a Godzilla hamster in a sombrero attacks Tokyo. “
When more people create and share DALL-E Mini images on Twitter and Redditand more and more new users arrived, Hugging Face found its server overloaded with traffic. “Our engineers didn’t sleep the first night,” Clément Delangue, CEO of Hugging Face, said on a video call from his home in Miami. “It’s really hard to service these models on a large scale; they had to fix everything”. In recent weeks, DALL-E Mini has delivered about 50,000 images per day.
Illustration: Staff with a WIRE / Face hugging
DALL-E Mini’s viral moment doesn’t just herald a new way to create memes. It also provides an early look at what could happen as AI tools help make-to-order images become widely available, and a reminder of uncertainties about the possible impact. of them. Algorithms that create custom artwork and photography can transform art and help businesses market, but they can also have the potential to manipulate and deceive. A warning on the DALL-E Mini website warns that it could “reinforce or exacerbate social prejudices” or “create prejudiced images against minorities.” “.
DALL-E Mini is inspired by a more powerful AI imaging engine called DALL-E (a compound word of Salvador Dali and WALL-E), revealed by AI research firm OpenAI in January 2021. DALL-E is more powerful but has not been made public, due to concerns that it will be misused.
It is common for breakthroughs in AI research to be quickly replicated elsewhere, often within a few months, and DALL-E is no exception. Boris Dayma, a machine learning consultant based in Houston, Texas, says he was intrigued by DALL-E’s original research paper. Although OpenAI did not release any code, he was able to beat the first version of DALL-E Mini at a hackathon organized by Hugging Face and Google in July 2021. produces low-quality images that are often barely noticeable, but Dayma has continued to improve it ever since. Last week, he changed the name of his project to Craiyon, after OpenAI’s request, he changed the name to avoid confusion with the original DALL-E project. The new site shows ads, and Dayma is also planning a premium version of its visual builder.
The image of the DALL-E Mini has a different look. Objects are often distorted and blurred, and people appear with faces or body parts missing or misaligned. But often being able to recognize what it’s trying to describe, and comparing the sometimes unedited output of the AI with the original prompt is often interesting.
