Hacker News Reader

Is there any reason we shouldn’t view diffusion models as any other tool? I can infringe copyright with photoshop too… even accidentally. If I generate original work with either that seems like fair game.

I imagine with the right prompt one could coax out a copywritten image even if it hadn’t ever seen it before

View on HN

There's still valid reason to be concerned here. For example, this implies building a diffusion model based on private data can leak it. If I can generate a whole bunch of prompts like "MRI Joe Biden brain tumor" and 1 out of a million times I get a consistent result, that's unacceptable.

Github Copilot could be leaking private code as well.

It seems the only way they can get an extraction is if a highly duplicated (vs the average) training image is used which is counter to a privacy concern.

This thread shows that there are outliers that are non-duplicated in the training set that still show up in results.

https://twitter.com/alexjc/status/1620466058565132288

Specifically, this post confirms cases of a single image in the training set being "memorized"

https://twitter.com/Eric_Wallace_/status/1620475626611421186...

Ah I see the section on this in the paper now, the 2nd half of 7.1 on page 14.

Reply on news.ycombinator.com