OpenAI is developing a new tool that will help users detect whether a photo was made by its DALL-E generative AI image generator and monitor the content created by the tech.
Additionally, the company is working on creating new watermarks on these AI-generated images to easily determine if they are from DALL-E, further helping to authenticate its works.
OpenAI: New DALL-E Image AI Detection Tool is Coming
(Photo: STEFANI REYNOLDS/AFP via Getty Images)
This tool will help researchers and the company verify the AI-generated content better, particularly if DALL-E 3 was behind the said work through the tool.
OpenAI said that this tool will help the world identify whether the content was made using its tech, in this case, DALL-E 3, which is powered by GPT-4.
The company is offering open access to its upcoming "image detection classifier" during a testing phase and asking for feedback from users. OpenAI said that this tool can identify if DALL-E made the image with up to 98 percent accuracy.
Conversely, this tool can only identify images made by other platforms with 5 to 10 percent accuracy.
Read Also: OpenAI's ChatGPT Can Now Use DALL-E AI Image Generation Directly on Chat
New Watermarks for DALL-E 3 On the Works
OpenAI also announced that it is joining the Coalition for Content Provenance and Authenticity (C2PA), with its investments marking a new era for its works.
With this, the company is working on new watermarks to add to their AI-generated works from images (via DALL-E), videos (via Sora), and audio.
The company claimed that it is adhering to C2PA's standards and looks to contribute to it, claiming that despite alterations to the image, it can still reveal the watermarks on it.
OpenAI and its DALL-E Generative AI
The latest version of OpenAI's renowned AI image generator, DALL-E 3, was released last year, offering the service directly from its platform and via its partner, Microsoft. On Microsoft's end, DALL-E 3 is directly accessible via Bing Chat, later to Copilot after its revamp, offering users a way to get their preferred image generations from the AI-powered search engine.
It is known that DALL-E 3 is also sporting the latest version of OpenAI's large-language model (LLM), GPT-4, offering this latest development to give users improved and more realistic photo generations.
OpenAI also promises that it will bring the best possible generation thanks to its most advanced GPT, especially with more details and specific to a user's prompt.
While there are safeguards made on DALL-E to prevent its misuse and create fake, malicious, and damaging photos from the platform, several bad actors still manipulate AI platforms to do their bidding.
That being said, OpenAI is taking a proactive approach to developing a new AI image detection tool to combat it. It is also introducing new watermarking methods that will allow everyone to determine if it was made by AI.
Related Article: Fighting Election Deepfakes: Microsoft, OpenAI Pump $2 Million Fund Aims to Counter AI Misinformation
