First Page Design
Site Theme
When OpenAI’s DALL-E 2 debuted on April 6, 2022, the idea that a computer could create photorealistic photographs on demand, based solely on textual descriptions, took many people by surprise. This launch ushered in an innovative and tumultuous era in the history of AI, marked by a sense of wonder and a polarizing moral debate that still resonates in the area of AI today.
Last week, OpenAI disabled the ability for new consumers to purchase building credits for the Internet edition of DALL-E 2, removing it. From a generational standpoint, it’s no surprise that OpenAI has recently started terminating the service. The old style of symbol generation was revolutionary for its time, but it has since been overtaken by the top point of detail of the DALL-E 3, and OpenAI has recently started implementing the editing features of the DALL-E 3.
But for a tight-knit group of artists and tech enthusiasts who showed up at the start of DALL-E 2, the service’s twilight marks the bittersweet end of an era in which the AI generation seemed shortly like a magical portal to infinite creativity. DALL-E 2 has been really mind-blowing,” illustrator Douglas Bonneville told Ars in an interview. “There was an exhilarating sense of unlimited freedom in those early days, which we all suspected AI was going to unleash. It was like a release from anything to anything else, but we never knew exactly what. “
Prior to DALL-E 2, AI symbol generation had been in the background for some time. Since the advent of computers with graphic screens in the 1950s, other people have been creating symbols with them. As early as the 1960s, artists such as Vera Molnar, Georg Nees, and Manfred Mohr let computers draw, creating works of art using generative algorithms. Experiments by artists such as Karl Sims in the 1990s led to one of the first introductions of neural networks into the process.
The use of AI in PC art resumed in 2015, when Google’s DeepDream used a convolutional neural network to bring psychedelic detail to existing symbols. Then there are turbines based on Transformer models, an architecture discovered in 2017 through an organization of Google researchers. OpenAI’s DALL-E1 debuted as a tech demo in early 2021 and Disco Diffusion was unveiled later that year. Despite those precursors, it could be said that DALL-E 2 marked the main breaking point for text-to-symbol generation, allowing the user to enter a description of what they wanted to see and have a corresponding symbol appear before their eyes.
When OpenAI first announced DALL-E 2 in April 2022, some corners of Twitter were temporarily filled with examples of surrealist art generated, such as teddy bears like mad scientists and astronauts on horseback. A lot of other people were really shocked. ” Ok, is that wrong?Tell me it’s wrong, the April Fool’s Day joke, joke a little late,” reads one of the first reactions on Twitter. “My brain can only fly so many times. I can’t take it anymore,” one Twitter user wrote in May.
More examples of DALL-E 2 art collected in threads soon followed, all from OpenAI and a carefully selected organization of two hundred beta testers.
When OpenAI started handing out those invitations for beta testing, a not-so-unusual bond temporarily gave rise to a small network of artists who felt like pioneers exploring the new generation together. “There was a crazy moment where some artists toyed with it. All friends,” said concept artist Danielle Baskin, who first won an invitation to use DALL-E 2 on March 30, 2022, and began testing in mid-April. “When I first accessed it, I felt like I had a portal to Infinite Worlds of Exchange. I didn’t think of it as “creating art”, it was like playing. I would stay up for hours just to explore. “
Because each DALL-E symbol was born from a spark written as “a photo of a statue gliding on ice” (based on learned associations between legends and symbols), the beta testers found that they fused language and their visual mind in new ways. . . “It was like getting carried away in a lab,” said an artist named Lapine in an interview with Ars. Lapine gained early access to DALL-E 2 on April 6 and began sharing his generations on Twitter. I was employing descriptive language in a way that I had never used before. “
Join the Ars Orbital Transmission email to receive weekly updates in your inbox. Register →