AI safety lacks a space to ask stupid or ballsy questions
(this page assumes that asking stupid or ballsy questions is important for learning/making intellectual progress)
I think the way that voting and crossposting work on LessWrong and the Alignment Forum significantly reduces the number of stupid or ballsy questions people can ask. I like asking these questions, but I feel like there's a ton of work involved in making these publicly acceptable. If I have 50 such questions (I'm sure I've written down at least 50 questions like this), I'm only posting like 5 of them on the internet, because it takes a lot of mental energy to simulate people's reactions, making it palatable would-be-downvoters, etc.