I’d like to distill AI Safety posts and papers, and I’d like to see more distillations generally. Ideally, posts and papers would meet the following criteria:
Potentially high-impact for more people to understand
Uses a lot of jargon or is generally complex and difficult to understand
Not as well-known as you think they should be (in the AI X-risk space)
What posts meet these criteria?
I think the longer blog posts by Anthropic and OpenAI on their approaches to alignment are very important, under-appreciated and sometimes (I think falsely) dismissed as disingenuous.
Commentary from skeptical researchers about these plans could be interesting to include as well.
I see you already volunteer on aisafety.info! From working on that project these are some areas I think could benefit from being made more accessible (on our platform or otherwise—we’re working on these but definitely could use the help + I would be really happy to see them worked on anywhere)
The research agendas and strategies of various alignment orgs and independent researchers
AI policy: The proposals that have been made and the content of active and forthcoming policy documents
Forecasting: The predictions that have been made and the methods by which we’ve come to them
Current models: What the most powerful models are doing and how
I’ll come back and add more if/when I think of more
I realize these are categories instead of specific documents but there’s just so much to be worked on! These are purely my views, and I haven’t run this past anyone else in the team, who I suspect have more thoughts. For anyone stumbling across this who’d like to help with the project but who isn’t familiar with it, we have a prioritized list of content we would like to cover on the site but don’t have yet.
I’m working on a related distillation project, I’d love to have a chat so we can coordinate our efforts! (riley@wor.land)