The latest AI image generation model Dall-E 3 is now integrated into OpenAI’s ChatGPT, greatly simplifying the process of creating text prompts to generate images. This is a significant step forward for OpenAI in the further development of its chatbot to make it even more versatile. According to OpenAI in a press release, compared to its predecessor, Dall-E 3 produces images that are not only more visually impressive, but also more detailed and sharper. In particular, improvements have been made to the display of text elements, hands and faces.
The latest incarnation of Dall-E 3 includes a “multi-level security system”. This system has been developed to ensure that “potentially harmful images” cannot be produced. This category includes photos that depict violence, indecency, hatred or sexual motifs. This system not only checks the generated images themselves, but also the requests entered by users to generate images.
ChatGPT users have the option to report problematic results or those that do not reflect their original input. Through this type of interaction, users can help to further improve and optimize the image generation model. The latest version of the image generator was already integrated into Bing, Microsoft’s search engine, at the beginning of October.
OpenAI claims to have taken precautions to prevent the AI image generator from being used to “create content that mimics the style of living artists, or images of well-known public figures”. In addition, a new internal tool has been developed that can recognize differences between images created by Dall-E 3 and those created elsewhere. This tool has proven to be extremely effective, with a hit rate of 99% for unmodified images and 95% for modified images. The main goal of this tool is to help people determine whether visual content has been generated using AI or not.
