The Path Ahead: Provenance Classifier
We’re actively researching and evaluating a provenance classifier—a powerful internal tool designed to determine whether an image was generated by DALL·E 3. In early internal testing, this classifier has demonstrated impressive accuracy, exceeding 99% for unmodified images and remaining over 95% accurate even with common modifications like cropping, resizing, or adding text to the generated image. While this tool is a promising step, it’s important to note that it can indicate the likelihood of an image being AI-generated but does not provide definitive conclusions.
The provenance classifier represents a part of the broader effort to help users identify AI-generated audio or visual content. This challenge requires collaboration across the AI ecosystem, including content distribution platforms. We’re committed to learning more about this tool’s effectiveness and its most useful applications and will continue to refine our approach.
Creative Controls
We’re attentive to creative and ethical considerations. DALL·E 3 has been designed to decline requests for images in the style of living artists, respecting their work and contributions. Additionally, we provide creators with the option to exclude their images from the training of our future image generation models, offering greater control and protection for their creative output.
In summary, DALL·E 3 in ChatGPT Plus and Enterprise represents a significant step forward in AI-powered creative content generation. With its creative potential, advanced safety measures, and commitment to ethical considerations, DALL·E 3 opens up new horizons for users, offering exciting possibilities for innovative projects and collaborations.
The future of AI-generated content is here, and we’re excited to embark on this journey with you.
