I guess I will copy my comment because I don't have change my mind about it: "So wait they only found 109 matches after generating 175 milion images using the prompt from the most duplicated samples from the dataset and SD v1.4? Also almost all of them have more than 300 copies in the dataset, so with a model with the same size and trained on a dedup dataset like SD 2.0/2.1 there will be almost no matches, even after generating 175 mln images and knowing the prompts used in the dataset. Finally Google at el need to explain how an attacker that want to extract images from a trained model somehow has the prompts for the top X duplicated images in the dataset but not the images themselves, and thus will going to spent an incredible amount of money to generate something like 175 mln samples and test them together to find the matches."
I also want to add that google seems to try really hard to show themselves as the good guys by not releasing its models because it's not safe enough, but in this paper they used an incredible amount of computation and show me otherwise.
Please don't copy/paste comments! Besides lowering signal/noise ratio, it makes merging threads a pain; it means we have to merge subthreads as well, to consolidate the replies your comment got.
Much better to just post a link, and ideally let us know about the split discussion at hn@ycombinator.com so we can merge.
7 comments
[ 0.19 ms ] story [ 29.4 ms ] thread[1] https://news.ycombinator.com/item?id=34596187
I guess I will copy my comment because I don't have change my mind about it: "So wait they only found 109 matches after generating 175 milion images using the prompt from the most duplicated samples from the dataset and SD v1.4? Also almost all of them have more than 300 copies in the dataset, so with a model with the same size and trained on a dedup dataset like SD 2.0/2.1 there will be almost no matches, even after generating 175 mln images and knowing the prompts used in the dataset. Finally Google at el need to explain how an attacker that want to extract images from a trained model somehow has the prompts for the top X duplicated images in the dataset but not the images themselves, and thus will going to spent an incredible amount of money to generate something like 175 mln samples and test them together to find the matches."
I also want to add that google seems to try really hard to show themselves as the good guys by not releasing its models because it's not safe enough, but in this paper they used an incredible amount of computation and show me otherwise.
Much better to just post a link, and ideally let us know about the split discussion at hn@ycombinator.com so we can merge.