OpenAI Releases Third Version of its Text-to-Image AI Image Tool DALL-E

DALL-E 3, a new version of OpenAI's AI image maker, has been released. This updated version is said to comprehend more subtleties and specifics in text cues, enabling users to produce more precise graphics. Because DALL-E 3 is built directly on ChatGPT, users may engage with the AI chatbot in conversation to create visuals and modify prompts.

OpenAI is now strengthening its services after introducing generative AI models that produce visuals based on text inputs supplied by users. The business that created ChatGPT is releasing DALL-E 3, the third iteration of its AI image maker.

What is new in DALL-E?

OpenAI claims that DALL-E 3 understands significantly more nuance and detail (in text prompts) than its previous systems. This, in turn, allows users to create images that more closely match users' ideas. "Modern text-to-image systems tend to ignore words or descriptions, forcing users to quickly learn to draw.

DALL-E 3 is a step forward in our ability to create images that match the text you provide," the company said. Sam Altman, CEO of OpenAI, shared examples of images created with DALL-E 3, highlighting the accuracy of results based on text instructions.

DALL-E 3 is designed for ChatGPT

OpenAI said DALL-E 3 is built natively on ChatGPT, meaning users can use the AI ​​chat to create images and refine testimonials – from a simple sentence to a detailed paragraph in the form of a conversation. "When you ask for an idea, ChatGPT automatically creates custom detailed invitations for DALL-E 3 to bring your idea to life. If you like a certain image but it's not quite right, you can ask ChatGPT to make changes with just a few words. ." the company noted.

DALL-E 3 Availability

DALL-E 3 will be available to ChatGPT Plus and Enterprise customers in early October. Images created with DALL-E 3 can be reprinted, sold or resold. OpenAI also said that DALL-E 3 does not yet produce violent, adult or hateful content. It has the flexibility to reject requests that ask for a public figure by name.

What is OpenAI Red Teaming Network?

In the meanwhile, as part of its continued endeavor to strengthen its AI systems, OpenAI recently introduced the OpenAI Red Teaming Network, a hired team of specialists to assist the business's AI model risk assessment and mitigation techniques.

As AI technologies, particularly generative technologies, become more widely used, red teaming is turning into an increasingly important phase in the building of AI models. Red teaming may identify biases in models like OpenAI's DALL-E 2, which has been found to magnify racial and sexual prejudices, and prompts that can make text-generating models, such as ChatGPT and GPT-4, disregard safety filters. However, red teaming does not always repair these biases.