Hacker News Reader

I'm still interested by other people's chats, so long as they are probing the limits of the model in ways I haven't thought of.

I don't want to see yet another conversation where someone asks it if it will become skynet or if it can write a haiku about whatever.

It can also be quite inspiring. If this thing makes you the next J.K. Rowling and in some shocking moment you reveal to your millions of fans that it wasn't actually you, it will be worth the hassle.

Why? We might learn about the failure modes of AI.

Can we arrive at some absolute rules about AI using the interactions?

At best you would get rules about this particular implementation and training approach. I don't see why we would expect those to apply to AI generally.