OpenAI has unveiled DALL-E 3, the latest version of its revolutionary AI system that creates images from text prompts. DALL-E 3 includes several new features aimed at improving the quality and safety of the generated images.
One of the most exciting new capabilities is integration with ChatGPT, OpenAI’s natural language chatbot. Users can now ask ChatGPT to generate detailed prompts for DALL-E 3, making it easier for novice users to create high-quality AI art. In demos, ChatGPT wrote prompts that resulted in imaginative images like mountain landscapes with ramen noodle snowcaps.
DALL-E 3 also has enhanced safety measures to prevent inappropriate content. OpenAI worked with external testers to identify vulnerabilities and trained the system to avoid generating lewd or hateful images. Specific public figures can no longer be depicted, and the model will decline attempts to mimic living artists’ styles.
The DALL-E system converts text prompts into photorealistic images using a complex AI technique called diffusion modeling. Since the original DALL-E launch in 2021, OpenAI has continuously refined the technology. DALL-E 3 is said to have much stronger contextual understanding compared to prior versions.
After facing criticism over potential misuse of earlier DALL-E systems, OpenAI will gradually roll out DALL-E 3 access starting with business users in October. The company has not yet announced plans for a free public release.
Rival AI art platforms from Stability AI, Midjourney and others have prompted legal action from artists concerned about copyright issues. In response, OpenAI will allow artists to opt-out of having their work copied by submitting examples for blacklisting.
While revolutionary, DALL-E 3 remains an imperfect work in progress. OpenAI acknowledges ongoing efforts will be needed to strengthen safety measures as the system continues evolving. But the enhanced capabilities of DALL-E 3 demonstrate the rapid pace of progress in AI generative art.
