How can one literally buy time (from x-risk) with money?

post by Alex_Altair · 2022-12-13T19:24:06.225Z · LW · GW · 3 comments

This is a question post.

Contents

3 comments

This post [LW · GW] talks about types of projects that can buy us more time to work on AI alignment. But like, I am just definitely not going to pivot to working on any of those projects right now. What pre-existing projects could someone donate money to, right now? (This seems especially relevant in the shadow of lost FTX funding.)

Answers

3 comments

Comments sorted by top scores.

comment by Chris_Leong · 2022-12-14T11:48:41.330Z · LW(p) · GW(p)

Excellent question. I think it's very hard to buy time as the main way to do this would be to persuade top researchers to stop pursuing capabilities research, but these people are very busy, hard to access, have high standards for arguments and have strong incentives to continue what they are doing.

Even though some of those interventions may buy time, the way that they do it seems to be rather indirect which makes me very uncertain of the extent to which they would buy us time. Part of me wonders whether "buying time" is even the correct frame for many of these interventions. For example, with demonstrating concerning capabilities, perhaps we should be thinking about these in terms of optionality value? The demonstration is sufficient to grow the movement somewhat, but it also increases the expected return on effort invested in movement-building.

Replies from: nem
comment by nem · 2022-12-14T16:40:02.803Z · LW(p) · GW(p)

Well for my own sanity, I am going to give money anyway. If there's really no differentiation between options, I'll just keep giving to Miri.

Replies from: Chris_Leong
comment by Chris_Leong · 2022-12-14T23:12:47.357Z · LW(p) · GW(p)

I'm not claiming that there's no differentiation between options - it's that you're buying movement growth (=extra research) + optionality (some of which may end up being converted into time).