✨
AI Summary
- Yann LeCun argues provably safe AI is impossible, like provably safe turbojets.
- Focus on AI systems that are un-jailbreakable by design, not patches.
- Objective-driven systems differ fundamentally from language models in their capabilities.
