AI Alignment

It feels like we're groping around to try to find where to start with AI alignment, which is a dangerous place to be when the tools are already the strong. I don't even know that I've seen a place where somebody's written down like, what is the target here? The closest thing weirdly, you know, I want to say fun, funnily enough, would be like Yidkowski's CEV 10 or 15 years ago.