Generative artificial intelligence (AI) is starting to be offered openly allowing anyone to generate images that have so far only been stored in the head due to lack of writing talent. The downside is that anyone can also produce fake images that look realistic. In a Tech Live event organized by WSJ, OpenAI CTO Mira Murati said they are developing a generative AI-generated image detector with 99% accuracy.
At the moment the high accuracy is only on the images generated by DALL.E 3 developed by them. Development is still underway but the date when it will be offered to the public has yet to be announced. Previously OpenAI also offered an essay detector generated by ChatGPT but discontinued it due to low detection rate.
The model used by generative AI is constantly changing, so the equipment that detects it also needs to be updated regularly to detect it. Initiatives like C2PA put Content Credentials icons on pictures with meta data on every AI-generated image but it still needs to be done.
Tags
APPS & GAMES