How To Extract ChatGPT Hidden Training Data | Making LLMs (e.g. Llama) Spill Out Their Training Data

What if ChatGPT can tell you more than it should? In this video, we unveil a remarkable attack on ChatGPT that exposes its hidden training data, complete with a live demonstration. We'll look at the accompanying blog post and arXiv paper, shedding light on the intricacies of this discovery, and explain the inner workings of how and why this attack is so effective in revealing hidden information.

Blog post: https://not-just-memorization.github.io/extracting-training-data-from-chatgpt.html
Paper: https://arxiv.org/abs/2311.17035

Discord: https://discord.gg/UaNPxVD6tv
Prepare for the Machine Learning interview: …

arxiv blog chatgpt data extract hidden light llama llms look making paper training training data video