OpenAI has announced the third version of its generative AI visual art platform DALL-E. This now allows users to create prompts using his ChatGPT and includes more safety options.
DALL-E converts text prompts to images. But even DALL-E 2 was wrong, often ignoring certain wording. OpenAI researchers say the latest version now has a better understanding of context.
A new feature in DALL-E 3 is integration with ChatGPT. ChatGPT eliminates the need for someone to come up with their own detailed information. Prompt to guide DALL-E 3. Just ask ChatGPT to create a prompt and the chatbot will write out a paragraph (DALL-E works better with longer sentences) for her DALL-E 3 to follow. Other users can use their own prompts if they have specific ideas for DALL-E.
In the demo, The Verge, Aditya Ramesh, principal researcher and head of the DALL-E team, asked ChatGPT to help him come up with a logo for a ramen restaurant in the mountains. ChatGPT then created a long prompt and DALL-E said he came up with four options. My favorite was the one depicting a snow-capped mountain of ramen, soup cascading down like a waterfall, and an egg pickled on the ground like a garden stone. However, it looked more like a nice product illustration than a traditional restaurant logo. According to OpenAI, the collaboration with this chatbot will allow more people to create his AI art, even if they are not good at coming up with prompts.
First released in January 2021, DALL-E was released ahead of other generative text-to-image AI art platforms by Stability AI and Midjourney. By the time DALL-E 2 is released in his 2022, OpenAI has announced that the platform will be closed following criticism that DALL-E can produce photorealistic explicit images and that bias can be seen when producing photos. We have established a waiting list to manage users. The company removed the waiting list and made DALL-E 2 available to the public last September.
This new version of DALL-E will be released first to ChatGPT Plus and ChatGPT Enterprise users in October, followed by laboratories and their API services in the fall. OpenAI plans to stagger the release of DALL-E 3, but has not said when a free public version will be released.
OpenAI claims that it has put significant effort into DALL-E 3 to create robust security measures to prevent the creation of obscene or potentially hateful images. OpenAI works with an external red team (a group that intentionally tries to break systems to test their safety) and has made sure that the language ignores certain words to avoid explicit or violent prompts. He said it relies on the input classifier, which is how the model is taught. DALL-E 3 also cannot recreate images of famous people if they are specifically named in the prompt.
Sandini Agarwal, a policy researcher at the company, said she had “high confidence” in the safety measures, but clarified that the model is constantly being improved and is not perfect. OpenAI representatives said in an email that DALL-E 3 has been trained to refuse to produce images in the style of living artists. Unlike his DALL-E 2, which can imitate art in the style of a particular artist, depending on the prompt.
OpenAI plans to allow artists to opt their art out of future versions of its text-to-image AI model, perhaps to avoid litigation. Creators can submit images to which they own the rights and request their removal using a form on the website. Future versions of DALL-E will allow you to block results that resemble an artist’s image or style. The artists sued DALL-E competitors Stability AI and Midjourney, as well as art website DeviantArt, for using their copyrighted work to train text-to-image models.