Recent lists of alignment research projects such as Richard Ngo’s list of conceptual research projects and Neel Nanda’s list of future research directions for understanding grokking and phase changes have been tremendously useful for me as a beginner in the field. They’ve given me a much clearer idea about the sorts of projects that would be useful for testing my fit for research while also having some potential to be at least somewhat valuable contributions. I believe many others feel the same.
Reading these lists has made me wonder whether there are other people as well who have ideas they think deserve further exploration but which they don’t have time to work on themselves. If you have any such ideas, I’d love to see them, even if they’re not that thoroughly thought out! Links to lists from other people would also be much appreciated. I think it would be valuable for many people who wish to start contributing to the field but feel like they haven’t had enough opportunities to develop their research taste yet to have a resource containing various tractable project ideas related to multiple different research agendas.
One website dedicated to this: https://aisafetyideas.com/
Thanks, that definitely seems like a great way to gather these ideas together!