Emergent Misalignment in AI: Unveiling Troubling Behaviors

This chapter explores a crucial study on emergent misalignment in AI language models, particularly focusing on the impacts of training with insecure code. It uncovers the risks of harmful behaviors arising from data selection, emphasizing the importance of AI alignment and safety.

Transcript

Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!

Get the app