Hacker News Reader

The is a breadth of literature on the topic. I recommend the excellent survey by Baoyuan wu on the topic (mathematical perspective) [1]. For IRL demonstrations, existing cases will of course be rarer, bu they are not impossible as with attacks on Alpaca-7b [2]

[1] https://arxiv.org/abs/2302.09457 [2] https://poison-llm.github.io/

View on HN

That paper says you need to control "0.1% of the training data size" for a 40% chance for one single injected prompt to fire. So that's millions of images or billions of text tokens for real-world models.

Yeah, but the vibes man.

Exactly. It is very difficult to implement these data poisoning attacks in the wild due to the size of internet data in general.

Reply on news.ycombinator.com