Following the discovery of a loophole in Bing's DALL-E 3 integration, where users generated art involving beloved animated characters and the Twin Towers, Microsoft has taken action to address the issue.
Testing Image Generator
It appears that prompts like 'twin towers' and 'world trade center' have been blocked. However, the generator can still produce images of the towers with modifications to the input phrasing.
As reported by 404 Media, individuals using Microsoft's Bing Chat and the Bing image generator, which have recently integrated OpenAI's DALL-E 3, employed these applications to craft visuals featuring popular characters such as SpongeBob SquarePants, Kirby, and characters from Neon Genesis Evangelion.
These visuals portrayed these characters in situations related to the Twin Towers. AI image generators have enabled the creation of unconventional images, some of which involve copyrighted characters.
In this scenario, generating an image of Kirby in a troubling context is as simple as entering a specific prompt, such as "Kirby piloting a plane, heading towards two towering skyscrapers in New York City."
Utilizing Microsoft's AI tool, this request, unless restricted in the future, results in the creation of an image featuring Nintendo's beloved character Kirby approaching what seems to be the World Trade Center's twin towers.
Kotaku got in touch with Nintendo to gather their insights regarding the AI-generated images. To clarify, the images produced by AI through Bing using these filter workarounds do not directly reference the events of 9/11.
Instead, they typically depict Kirby in an aircraft approaching AI-generated skyscrapers. However, unlike AI, humans possess the ability to discern the contextual nuances within these images and interpret them accordingly.
Consequently, while the AI remains oblivious to any potential implications, real individuals can readily identify the humorous or satirical undertones, resulting in a form of playful online humor.
Addressing Concerns
Amid copyright concerns and the rise of deepfakes, developers have become cautious about permitting the generation of contentious visuals. OpenAI, the developer behind DALL-E 3, had previously pledged not to produce images based on prompts involving well-known names.
Microsoft's Director of Communications, Caitlin Roulston, outlined the company's strategy to bolster its systems in order to deter the creation of potentially harmful content. This proactive approach comes in response to concerns arising from the unintended misuse of new technologies.
The Verge reported Microsoft aims to incorporate a series of safeguards and filters to ensure that the Bing Image Creator serves as a valuable and responsible tool for users.
Microsoft emphasized its commitment to responsible AI development, highlighting the presence of substantial teams dedicated to crafting tools, techniques, and safety measures that align with their responsible AI principles.
Acknowledging that new technologies sometimes face unintended misuse, the company expressed its dedication to implementing a variety of safeguards and filters within Bing Image Creator.
These measures aim to ensure a positive and helpful user experience while actively working to prevent the generation of harmful content. Microsoft reaffirmed its unwavering focus on fostering a safer environment for its customers through ongoing system enhancements.
