AI models spit out photos of real people and copyrighted images

Stable Diffusion is open source, which means anyone can analyze and study it. Imagen is closed, but Google has granted researchers access. Singh says the work is a great example of the importance of giving research access to these analytics models, and he argues that companies should be equally transparent with other AI models, such as ChatGPT. from OpenAI.

However, while the results are impressive, they come with some caveats. The images the researchers managed to extract either appeared multiple times in the training data or were highly unusual compared to other images in the dataset, says Florian Tramèr, assistant professor of computer science at ETH Zürich, who was part of the group.

People who look unusual or have unusual names are at a higher risk of being remembered, Tramèr says.

The researchers were only able to extract relatively few exact copies of individuals’ photos from the AI ​​model: only one in a million images was a copy, according to Webster.

But it’s still worrying, says Tramèr: “I really hope nobody looks at these results and says, ‘Oh, actually, these numbers aren’t that bad if it’s just one in a million.’”

“The fact that they are greater than zero is what matters,” he adds.