 |
| Image: Medium |
OpenAI has expanded access to its cutting-edge text-to-image generator, DALL-E 3, offering it to ChatGPT Plus and Enterprise users within the ChatGPT app. This move comes after OpenAI's efforts to ensure safety and reliability in the model's use, enabling a broader release.
DALL-E 3, introduced just a month ago, builds on the previous DALL-E 2, empowering users to craft more detailed and extensive prompts for image generation, thanks to its integration with ChatGPT. Remarkably, Microsoft's Bing Chat and Bing Image Generator became the first platforms to provide public access to DALL-E 3, even before ChatGPT.
Despite its potential, concerns about harmful content have surfaced. Users have playfully created images of famous landmarks morphed into whimsical characters and even engaged in risky imagery. Microsoft took steps to block some prompts, yet users found workarounds to produce similar outcomes.
Text-to-image generators like Midjourney, Stable Diffusion, and earlier DALL-E versions have faced controversies, generating copyrighted materials, non-consensual content, and distortions of subjects' ethnicity and public figures.
OpenAI is now emphasizing extensive measures and research to enhance the safety of DALL-E 3. They are determined to limit content resembling living artists' work, images of public figures, and enhance demographic diversity in generated images. Additionally, OpenAI employs an "provenance classifier" tool with a remarkable 99% accuracy in detecting DALL-E 3-generated images, further reinforcing their commitment to responsible AI development.
Source:
www.theverge.com
