Update for me: I think we (in the AI risk/rationalist/alignment community) should try especially hard now to actually be nice.

If you've been in this for some time, it's incredibly easy to become a bit jaded, because you likely had terrible arguments thrown at you for years.

But more mainstream people who will now first hear about the concerns (about transormative AI) will often be more open to the possibility that this is a huge problem than we might intuitively think.

And they *really* don't like it if they feel like they're being talked down to.

To be clear: I also often fail at this! This is why it's an update for me. I will invest to become better at this.

I think what can often happen is that the person who is new to this will think “why does the AI risk person seem suspicious that I don't 'get' the concerns. I do!“

But it's just an adaptive pattern that has been trained *for years* inside the emotional brain of the AI risk person. We've talked to a lot of bad faith people.

From now on, there will be *many* more people who we've never talked to before about this. We need to act like it!

(Also, I'm not saying everybody here is bad at this! Obviously not! A lot of people are fantastic.

But I think pushing us towards “being nicer“ on the margins could, at this moment, be incredibly valuable and I think it's currently massively underrated.)

(And also obviously: Making good arguments and saying true things is still almost always more important than being nice.)

If we're nice and polite — and have a good map of the territory — the world will be right.