Today, DALL-E 2, OpenAI’s AI system that can generate images on demand or edit and refine existing images, is becoming more widely available. The company announced in a blog post that it will accelerate access for waitlisted customers with the goal of reaching approximately 1 million people in the next few weeks.
With this “beta” launch, DALL-E 2, which was free to use, will switch to a credit-based fee structure. First-time users will receive a limited amount of credits that can be used to generate or edit an image or create an image variant. (Generations return four images, while edits and variations return three.) Credits will replenish each month up to 50 in the first month and 15 per month thereafter, or users can purchase additional credits in $15 increments.
Here’s a chart with the specifics:
Image credits: OpenAI
Artists in need of financial assistance will be able to apply for subsidized access, OpenAI says.
DALL-E’s successor, DALL-E 2 was announced in April and became available to a select group of users earlier this year, recently passing the 100,000 user mark. OpenAI says the wider access is made possible by new approaches to mitigating bias and toxicity in DALL-E 2 generations, as well as an evolution in the policy-driving images produced by the system.
An example of the types of images DALL-E 2 can generate. Image credits: OpenAI
For example, OpenAI said this week it implemented a technique that encourages DALL-E 2 to generate images of people that “more accurately reflect the diversity of the world’s population” when given a prompt describing a person of unspecified race or gender. The company also said it now rejects uploads of images containing realistic faces and attempts to create the likeness of public figures, including prominent political figures and celebrities, while improving the accuracy of its content filters.
Generally speaking, OpenAI does not allow DALL-E 2 to be used to create images that are not “Rated G” or that could “cause harm” (eg images of self-harm, hate symbols, or illegal activity) . It also previously prohibited the commercial use of generated images. Starting today, however, OpenAI is giving users “full usage rights” to commercialize the images they create with DALL-E 2, including the right to reprint, sell, and merchandise — including images they generated during the early preview.
As demonstrated by DALL-E 2 derivatives such as Craiyon (formerly DALL-E mini) and the unfiltered DALL-E 2 itself, image generation AI can very easily recognize the biases and toxicity embedded in the millions of images from the network in use to train them. Futurism managed to prompt Craiyon to create images of burning crosses and Ku Klux Klan rallies, and found that the system made racist assumptions about identities based on “ethnic-sounding” names. OpenAI researchers note in an academic paper that the open-source DALL-E implementation can be trained to make stereotypical associations such as generating images of white men in business suits for terms like “CEO.”
While the OpenAI-hosted version of DALL-E 2 was trained on a dataset filtered to remove images that contained obvious violent, sexual, or hateful content, the filtering has its limits. Google recently said it would not release an AI generation model it developed, Imagen, due to risks of abuse. Meanwhile, Meta has limited access to Make-A-Scene, its art-focused image generation system, to “prominent AI artists.”
OpenAI emphasizes that the hosted DALL-E 2 includes other safeguards, including “automated and human monitoring systems” to prevent things like the model from remembering faces that appear frequently on the Internet. Still, the company admits it still has work to do.
“Expanding access is an important part of our responsible deployment of AI systems because it allows us to learn more about real-world usage and continue to iterate our safety systems,” OpenAI wrote in a blog post. “We continue to explore how AI systems like DALL-E can reflect biases in their training data and different ways we can deal with them.”
Add Comment