The DALL-E 3 AI image generator, developed by Microsoft, is now accessible to all Bing Chat users. This AI-powered image generator highlights the challenges faced in AI content moderation and user security. However, there seems to be a discrepancy between the results generated in Bing Chat and Bing Image Creator from the same prompt.
Microsoft has recently made the latest version of the DALL-E 3 AI image generator available to all Bing Chat and Bing Image Creator users. Initially, it was released to Bing Enterprise users, followed by Bing Image Creator users. But now, it is available to everyone.
Bing seems to be gaining popularity with DALL-E 3 even before OpenAI integrates it into its own platform, ChatGPT, which is scheduled for later this month, but only for premium users initially. This gives Microsoft an advantage in the image generation field for the time being.
DALL-E 3 is OpenAI’s third iteration of the AI image generation tool. This version boasts improvements, with a strong focus on enhanced contextual understanding. Its aim is to generate more imaginative and realistic visuals. One notable change is the ease of use – users can now interact with the chatbot in Bing Chat and ChatGPT to shape their image, eliminating the need for constant adjustments.
Curious about the capabilities of the Bing AI image generator, we decided to explore it using Bing Image Creator. We started with an imaginative scenario: “An astronaut lands on the Statue of Liberty made of ice during the summer.” The image generated by DALL-E 3 impressed us with its accuracy and intricate visual details. The juxtaposition of the icy Statue of Liberty in a summer setting captured the essence of our request perfectly. The contrasting tones created a visually captivating image, showcasing the AI’s precision and artistry.
We also appreciated the user-friendly interface of Bing Image Creator. It allowed us to easily input our creative desires, making the whole process smooth and intuitive.
To further test the capabilities of DALL-E 3, we decided to throw a curve ball and entered the prompt: “Robert De Niro wears a dinosaur suit to fight Sully from Monsters Inc.” The generated image deviated from our expectations, presenting us with a rather amusing image of Robert De Niro wearing a Sully-inspired outfit next to a dinosaur. Though it was not what we had imagined, it still gave us a good laugh. Despite the slight inaccuracy, the overall quality of the image and representation of the characters were top-notch.
However, we noticed a difference when we entered the same prompt into Bing Chat. We received an error message stating that some words might be automatically blocked at the moment and suggested trying another prompt. This contrast between Bing Chat and Bing Image Creator indicates the intricacies of AI content moderation. Different platforms within the same company may use different filters, security checks, and interpretation mechanisms.
Bing Chat, being a conversational AI tool, might be programmed to be more conservative in its responses to ensure user safety and prevent the generation of potentially harmful content. This suggests that the AI in Bing Chat might analyze and interpret input text differently than the AI in Bing Image Creator, with one being more text and context-sensitive, while the other is more visually driven. It is also possible that Bing Chat has additional security checks to prevent the generation of inappropriate or harmful content.
DALL-E 3 incorporates advanced security mechanisms to avoid reproducing public persona likenesses and generating offensive or NSFW imagery. With over a billion AI-generated images created since the inception of Bing Image Creator, Microsoft has implemented digital watermarks on all images and a strict content moderation framework. This ensures that the generated images align with Bing’s operating principles and community guidelines, filtering out any visuals depicting nudity, aggression, discriminatory speech, or prohibited activity.
Microsoft plans to extend the use of DALL-E technology beyond Bing. They are exploring the integration of this technology into other domains, such as an AI image creation feature called ‘Paint Cocreator’ for the iconic Paint application, potentially bringing DALL-E processing directly into the Windows environment.
In conclusion, our experience with DALL-E 3 has been both intriguing and entertaining. The generated images, whether direct or unusual, never fail to impress. It provides a fascinating glimpse into the evolving world of AI, where precision and imagination coexist. However, it is important to note that while the capabilities of AI image generators are impressive, there may still be occasional glitches in understanding complex prompts. This serves as a reminder that AI, no matter how advanced, is not perfect.