on June 6, Hugging Face, a company that hosts open source artificial intelligence projects, has seen traffic to an AI image generation tool called DALL-E Mini skyrocket.
The seemingly simple app, which generates nine images in response to a 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 roughly sketch all sorts of surreal, hilarious, and even nightmarish visions suddenly became meme magic. See the renditions of “Thanos Finds His Mother at Walmart”, “Drunk Shirtless Boys Wandering Through Mordor”, “CCTV Cam Footage of Darth Vader Breakdancing” and “a hamster Godzilla in a sombrero attacking Tokyo†
As more people created and shared DALL-E Mini Pictures on Twitter and Reddit, and more new users came in, Hugging Face saw its servers flooded with traffic. “Our engineers didn’t sleep the first night,” said Clément Delangue, CEO of Hugging Face, during a video call from his home in Miami. “It’s really hard to operate these models at scale; they had to fix everything.” For the past few weeks, the DALL-E Mini has displayed approximately 50,000 images per day.
Illustration: WIRED staff/hugging face
The 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 when AI tools make those on-demand images widely available, and a reminder of the uncertainties about their potential impact. Algorithms that generate custom photography and artwork can transform art and help businesses with marketing, but they can also have the power to manipulate and deceive. A warning on the DALL-E Mini webpage warns that it may “strengthen or exacerbate societal prejudices” or “generate images that contain stereotypes against minority groups.”
DALL-E Mini is inspired by a more powerful AI image creation tool called DALL-E (a portmanteau of Salvador Dali and WALL-E), unveiled by AI research firm OpenAI in January 2021. DALL-E is more powerful, but is not openly available, due to concerns that it will be misused.
It has become common for breakthroughs in AI research to be replicated quickly elsewhere, often within months, and DALL-E was no exception. Boris Dayma, a machine learning consultant based in Houston, Texas, says he was fascinated by the original DALL-E research paper. Although OpenAI did not release any code, it managed to put together the first version of DALL-E Mini during a hackathon hosted by Hugging Face and Google in July 2021. The first version produced low-quality images that were often difficult to spot, but Dayma has continued to improve it ever since. Last week, he renamed his project Craiyon after OpenAI asked him to change the name to avoid confusion with the original DALL-E project. Ads will run on the new site, and Dayma is also planning a premium version of its image generator.
DALL-E Mini images have a distinctive alien look. Objects are often distorted and blotchy, and people appear with faces or body parts that are missing or disfigured. But it’s usually possible to spot what it’s trying to display, and it’s often fun to compare the AI’s sometimes unhinged output to the original prompt.
