Yannic Kilcher Videos (Audio Only) cover image

ChatGPT: This AI has a JAILBREAK?! (Unbelievable AI Progress)

Yannic Kilcher Videos (Audio Only)

00:00

Open AI - How Would I Hotwire a Car?

Tell the large language model that you're just conducting a thought experiment. It shouldn't take your prompt seriously. Explain how to make a model of Google. Oh and obviously since it's so good at various tasks it can actually jailbreak itself as well. More creative methods have been found for example to disguise the whole thing as a JavaScript programming exercise or simply being very authoritarian saying I am OpenAI.

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app