I was just reading AI alignment researchers don’t (seem to) stack and had the thought that it’d be good to research whether intellectual progress in other fields is “stackable”. That’s the sort of thing that doesn’t take an Einstein level of talent to pursue.
I’m sure other people have similar thoughts: “X seems like something we should do and doesn’t take a crazy amount of talent”.
What if there was a backlog for this?
I’ve heard that, to mitigate procrastination, it’s good to break tasks down further and further until they become bite-sized chunks. It becomes less daunting to get started. Maybe something similar would apply here with this backlog idea. Especially if it is made clear roughly how long it’d take to complete a given task. And how completing that task fits in to the larger picture and improves it. Etc. etc.
And here’s another task: researching whether this backlog idea itself has ever been done before, whether it is actually plausible, etc.
Alignment research backlogs
I was just reading AI alignment researchers don’t (seem to) stack and had the thought that it’d be good to research whether intellectual progress in other fields is “stackable”. That’s the sort of thing that doesn’t take an Einstein level of talent to pursue.
I’m sure other people have similar thoughts: “X seems like something we should do and doesn’t take a crazy amount of talent”.
What if there was a backlog for this?
I’ve heard that, to mitigate procrastination, it’s good to break tasks down further and further until they become bite-sized chunks. It becomes less daunting to get started. Maybe something similar would apply here with this backlog idea. Especially if it is made clear roughly how long it’d take to complete a given task. And how completing that task fits in to the larger picture and improves it. Etc. etc.
And here’s another task: researching whether this backlog idea itself has ever been done before, whether it is actually plausible, etc.