What will I do to make AI Impacts dinner better tonight, and will it work?
resolved Nov 18
Distribute name tags
Streamlined setup for hot toddies, better
Invite at least 15 people from Rethink Priorities's longtermism/AI teams, better
Bring inflatable sofa outside, better
Provide a larger pile of blankets, better
Invite at least 15 people from Rethink Priorities's longtermism/AI teams
No conversations bigger than 3, better

Will it work = do I guess it made the event better or worse afterwards?

"Provide a larger pile of blankets, better"

"Provide a larger pile of blankets, worse"
"Give people name tags, worse"
"Give people labels on which to write a thing they are interested in re the future of AI, better"

"Make another prediction market about what to talk to people about and then adhere to its suggestions, worse"
"Try to start an argument about P(doom), better"

This is the event: https://www.facebook.com/events/5790952050961422 or https://partiful.com/e/xFJdBPZLR780r307Pp4o

Last I checked it had ~34 RSVPs yes, and ~50 'interested'.

Betterness is in terms of something like the impartial good, not my own experience, relative to the counterfactual where I didn't do the thing. (Though that does include my own experience, so if a thing is too much effort to be worth it, that counts as worse.)

Things done to make it better have to be things I didn't do last time (e.g. 'buy pizza, better' won't be marked successful, because I did it last time).

If I can't tell if I did a thing, or whether it was good, I'll take my best guess. If a thing is an unfulfilled conditional, I will mark it unsuccessful though logically it is successful (e.g. "if there is a flying pig, try to catch it, better" will be marked unsuccessful).

I can't tell whether I'll be able to choose multiple correct answers, so if not, I'll choose the thing with the largest absolute effect on goodness (positive or negative). [ETA: seems like I can]

Get Ṁ600 play money
Sort by:

I counted 'distribute name tags' even though Austin did it not literally me, because I did give permission and thus control whether it happened. I also counted 'Invite at least 15 people from Rethink Priorities's longtermism/AI teams' even though I had already done it at the time of the market. I personally enjoyed the streamlined hot toddy situation enough to think it probably worthwhile assuming a small number of others are like me, but I didn't have that much evidence about it. I didn't do the other things, but had my eye on the blanket situation and might have got more if the initial pile ran out, so maybe that made the party better in expectation, but not in reality.