Did you only get papers or also lesswrong posts? there are a lot of very high quality lesswrong posts. I’ve been thinking about doing something similar to this, would be mainly focusing on lesswrong posts by default. Interested in reusing your list of papers, likely I’ll process them a bit differently if I get to this
edit: probably good to list some authors I’d want to make sure to get:
Scott Garrabrant
Abram Demski
Vanessa Kosoy
Richard Ngo
John Wentworth
Wei Dai
Andrew Critch
Steven Byrnes
TurnTrout
Cleo Nardo
Diffractor
Jan Kulveit
Jacob Hilton?
Cole Wyeth?
plex?
Zach Furman
J Bostock
Quinn
Max von Hippel
Raemon?
davidad?
Jeremy Gillen
List Is Incomplete Because I Was Reciting From Memory And Then Scrolling The Lesswrong Home Page And Definitely Missed At Least One Very Important Name But I Am Not At All Intending To Include Everyone Who Posts On Less Wrong
There are many more authors who have written posts that I think are good the way these posters’ posts are good, but who are not prolific, and so I don’t have their names memorized. There are also many people who are not working on alignment that is intended to take scratches at the overwhelming-superintelligence-alignment problem, only on local alignment, which is merely hoped to be a useful tool in climbing the overwhelming-superintelligence wall; they might be worth including, but it would be important to warn Claude that local alignment and asymptotic alignment are different things (I have a post upcoming about this).
The original search before I went with the references based approach got a couple posts, but I think clearly not enough. I couldn’t figure out a good systematic way to get posts, but I will definitely spend some more time thinking about this and add a tag for LW/AI Alignment posts,
https://github.com/StampyAI/alignment-research-dataset needs work but is the codebase I’ll be improving when I get back to this. I’d encourage you to steal from it. perhaps clone it in ../ and tell claude code to look at it as needed. note: a major todo for me is getting it to get comments, which it doesn’t do now.
edit: I can also send you a messier codebase that is able to get full user data. also load up Wei Dai’s updated userscript, I think it has the important access patterns you need for this
“I’d encourage you to steal from it. perhaps clone it in ../ and tell claude code to look at it as needed. note: a major todo for me is getting it to get comments, which it doesn’t do now.”
Did you only get papers or also lesswrong posts? there are a lot of very high quality lesswrong posts. I’ve been thinking about doing something similar to this, would be mainly focusing on lesswrong posts by default. Interested in reusing your list of papers, likely I’ll process them a bit differently if I get to this
edit: probably good to list some authors I’d want to make sure to get:
Scott Garrabrant
Abram Demski
Vanessa Kosoy
Richard Ngo
John Wentworth
Wei Dai
Andrew Critch
Steven Byrnes
TurnTrout
Cleo Nardo
Diffractor
Jan Kulveit
Jacob Hilton?
Cole Wyeth?
plex?
Zach Furman
J Bostock
Quinn
Max von Hippel
Raemon?
davidad?
Jeremy Gillen
List Is Incomplete Because I Was Reciting From Memory And Then Scrolling The Lesswrong Home Page And Definitely Missed At Least One Very Important Name But I Am Not At All Intending To Include Everyone Who Posts On Less Wrong
There are many more authors who have written posts that I think are good the way these posters’ posts are good, but who are not prolific, and so I don’t have their names memorized. There are also many people who are not working on alignment that is intended to take scratches at the overwhelming-superintelligence-alignment problem, only on local alignment, which is merely hoped to be a useful tool in climbing the overwhelming-superintelligence wall; they might be worth including, but it would be important to warn Claude that local alignment and asymptotic alignment are different things (I have a post upcoming about this).
The original search before I went with the references based approach got a couple posts, but I think clearly not enough. I couldn’t figure out a good systematic way to get posts, but I will definitely spend some more time thinking about this and add a tag for LW/AI Alignment posts,
https://github.com/StampyAI/alignment-research-dataset needs work but is the codebase I’ll be improving when I get back to this. I’d encourage you to steal from it. perhaps clone it in ../ and tell claude code to look at it as needed. note: a major todo for me is getting it to get comments, which it doesn’t do now.
edit: I can also send you a messier codebase that is able to get full user data. also load up Wei Dai’s updated userscript, I think it has the important access patterns you need for this
Okay, I finished the first pass at this: https://sladebyrd.com/ai-safety-db/posts
any thoughts?
“I’d encourage you to steal from it. perhaps clone it in ../ and tell claude code to look at it as needed. note: a major todo for me is getting it to get comments, which it doesn’t do now.”
Working on this right now