Finally, DALL-E 2, OpenAI’s AI image generation system, is available as an API, meaning developers can integrate the system into their apps, websites and services. In a blog post today, OpenAI announced that any developer can start harnessing the power of DALL-E 2, which more than three million people are now using to produce over four million images per day, once an account is created. OpenAI API as part of the public beta.
The price of the DALL-E 2 API varies by resolution. For 1024 × 1024 images, the cost is $ 0.02 per image; 512 × 512 images cost $ 0.018 per image; and 256 × 256 images is $ 0.016 per image. Volume discounts are available to companies working with the OpenAI business team.
As with the beta version of DALL-E 2, the API will allow users to generate new images from text messages (for example “a fluffy bunny jumping through a field of flowers”) or to edit existing images. Microsoft, a close partner of OpenAI, is leveraging this in Bing and Microsoft Edge with its Image Creator tool, which allows users to create images if web results don’t return what they’re looking for. The CALA fashion design app uses the DALL-E 2 API for a tool that allows customers to refine design ideas from text descriptions or images, while launching Mixtiles photos is leading it to a stream of creating works of art for its users.
Not much in the way of policy is changing with the launch of the API, which risks disappointing those who fear that generative AI systems like DALL-E 2 will be released without sufficient consideration for the ethical and legal issues they pose. As before, users are bound by OpenAI’s terms of service, which prohibits the use of DALL-E 2 to generate overtly violent, sexual or hateful content. OpenAI also continues to prevent users from uploading images of people without their consent or images they do not have the rights to, employing a mix of automated and human monitoring systems to enforce this.
A small change is that images generated with the API should not contain a watermark. OpenAI introduced the watermark during the DALL-E 2 beta as a way to indicate which images originated from the system, but chose to make it optional with the launch of the API.
“We encourage developers to disclose that the images are generated by artificial intelligence, but they do not need to include the DALL-E 2 signature,” Luke Miller, OpenAI product manager overseeing the development of DALL-E 2, told TechCrunch. .
OpenAI also uses prompt and image level filters with DALL-E 2, although the filters complained by some customers are overly zealous and inaccurate. And the company has focused a portion of its research efforts on diversifying the types of images the DALL-E 2 generates, with the goal of fighting the prejudices that text-to-image AI systems are known to fall victim to. (eg mainly generating images of white men when prompted with text such as “CEO examples”).
But these steps have not dispelled all the critics. In August, Getty Images banned the uploading and sale of illustrations generated using DALL-E 2 and other similar tools, following similar decisions by sites like Newgrounds, PurplePort, and FurAffinity. Getty Images CEO Craig Peters told The Verge that the ban was caused by concerns about “correct issues not addressed,” as training datasets for systems like DALL-E 2 contain copyrighted images taken from the web.
Trying to find a middle ground, Getty Images rival Shutterstock recently announced it would start using DALL-E 2 to generate content, but at the same time launch a “grant fund” to reimburse creators when the company was selling work. to train text-to-image AI systems. It is also banning third-party uploaded AI art to minimize the potential for copyrighted work to make its way onto the platform.
Technologists Mat Dryhurst and Holly Herndon are leading an effort called Source + to allow people to ban the use of their work or likeness for AI training purposes. But it is voluntary. OpenAI hasn’t said if it will participate, nor if it will ever introduce a self-service tool to allow rights holders to exclude their work from training or content generation.
In an interview, Miller revealed little in the way of details about the new mitigating measures, other than the fact that OpenAI has improved its techniques to prevent the system from generating distorted, toxic, and otherwise offensive content that customers might find objectionable. You described the open API beta as an “iterative” process, which will involve working with “users and artists” over the next few months as OpenAI scales the infrastructure that powers DALL-E 2.
Of course, if the beta of DALL-E 2 is any indication, the API program will evolve over time. At first, OpenAI disabled the ability to edit people’s faces with DALL-E 2, but later enabled the feature after making improvements to its security system.
“We’ve worked a lot on this side of things, both through the images you upload and the prompts you send, to align them with our content policy and come up with several mitigations to filter at the prompt level and at the image level to make sure that is in line with our content policy. So, for example, if someone uploads an image that contains symbols of hate or blood, as very, very, very violent content, they would be rejected, “Miller said. “We always think about how we can improve the system.”
But while OpenAI seems eager to avoid the controversy surrounding Stable Diffusion, the open source equivalent of DALL-E 2 that was used to create pornographic, gore and celebrity deepfakes, it leaves API users the choice of exactly how and where to distribute. its technology. Some, like Microsoft, will undoubtedly take a measured approach, slowly launching products based on DALL-E 2 to gather feedback. Others will dive headfirst, embracing both technology and the ethical dilemmas that come with it.
If one thing is certain, it is that there is a pent-up demand for generative AI: to hell with the consequences. Even before the API was officially available, developers were publishing workarounds to integrate DALL-E 2 into apps, services, websites, and even video games. With the launch of the public beta, fueled by the formidable marketing force of OpenAI, synthetic images are ready to truly enter the mainstream.