Opinions expressed are my own and not endorsed by anyone.
Formerly @ ARC Evals aka METR
Posts
Wiki Contributions
Comments
What do you mean by "following through"? Just sending another email?
(My track record of 0% accuracy on which messages will politically snowball is holding up very well. I'm glad that sometimes people like you say things the way you say them, rather than only people like me saying things how I say them.)
I wonder if a chat loop like this would be effective at shortcutting years of confused effort maybe in research andor engineering. (The AI just asks the questions and the person answers.)
- "what are you seeking?"
- "ok how will you do it?"
- "think of five different ways to do that"
- "describe a consistent picture of the consequences of that"
- "how could you do that in a day instead of a year"
- "give me five very different alternate theories of how the underlying system works"
Questions like that can be surprisingly easy to answer. Just hard to remember to ask.
I would add one. I haven't found a compelling thing to aim for long term. I have asked many people to describe a coherent positive future involving AI. I have heard no good answers. I have been unable to produce one myself.
Are we playing a game that has no happy endings? I hope we are not.
I swear to never joke again sir
1I assumed somebody had. Maybe everyone did haha
From the frontpage:
https://www.lesswrong.com/posts/zAqqeXcau9y2yiJdi/can-we-build-a-better-public-doublecrux
https://www.lesswrong.com/posts/bkr9BozFuh7ytiwbK/my-hour-of-memoryless-lucidity
https://www.lesswrong.com/posts/ANGmJnZL2fskHX6tj/dyslucksia
Like all of them basically.
most of the value is in even figuring out how to diagram the posts
Think of it like a TLDR. There are many ways to TLDR but any method that's not terrible is fantastic
His sister's accusations that he blocked her from parent's inheritance and that he molested her when he was a young teenager and that he got her social media accounts flagged as spam to hide the accusations