Interesting ideaâthe implementation seems a bit off to me though, e.g. the first prompt I tried was one of the pre-written ones, and it said âgreat questionsâ plural even though I only asked one, and then it answered three different questions in one long message.
I think i tis more natural (and I would assume more effective) to have more of a back and forth.
Also, this is probably onbious to you, but just because someone gives a 10â10 of course doesnât mean they have actually been persuaded or will change their actions (though hopefully for some people it does, and what I like about this intervention is how cheap/âscalable it is).
Yeah, the chatbot also gives a reply to âWhy do they think that? Why care about AI risk?â, which is a UX problem, it hasnât been a priority.
Thatâs true, but the scale shows âcompletely changed my mindâ at the right side + people say stuff in the free-form section, so Iâm optimistic that people do change their minds.
Some people say 0â10 because theyâve already been convinced. (And we have a couple of positive response from AI safety researchers, which is also sus, because presumably, they wouldnât have changed their mind.) People on LW suggested some potentially better questions to ask, weâll experiment with those.
Iâm mostly concerned about selection effects: people who rate the response at all might not be a representative selection of everyone who interacts with the tool.
Itâs effective if people state their actual reasons for disagreeing that AI would kill everyone, if made with anything like the current tech.
Interesting ideaâthe implementation seems a bit off to me though, e.g. the first prompt I tried was one of the pre-written ones, and it said âgreat questionsâ plural even though I only asked one, and then it answered three different questions in one long message.
I think i tis more natural (and I would assume more effective) to have more of a back and forth.
Also, this is probably onbious to you, but just because someone gives a 10â10 of course doesnât mean they have actually been persuaded or will change their actions (though hopefully for some people it does, and what I like about this intervention is how cheap/âscalable it is).
Yeah, the chatbot also gives a reply to âWhy do they think that? Why care about AI risk?â, which is a UX problem, it hasnât been a priority.
Thatâs true, but the scale shows âcompletely changed my mindâ at the right side + people say stuff in the free-form section, so Iâm optimistic that people do change their minds.
Some people say 0â10 because theyâve already been convinced. (And we have a couple of positive response from AI safety researchers, which is also sus, because presumably, they wouldnât have changed their mind.) People on LW suggested some potentially better questions to ask, weâll experiment with those.
Iâm mostly concerned about selection effects: people who rate the response at all might not be a representative selection of everyone who interacts with the tool.
Itâs effective if people state their actual reasons for disagreeing that AI would kill everyone, if made with anything like the current tech.