I am very excited about finding scalable ways to collect large volumes of high-quality data on weird, specific tasks. This seems very robustly useful for alignment, and not something we’re currently that good at. I’m a bit less convinced that this task itself is particularly useful.
Have you reached out to e.g. https://www.surgehq.ai/ or another one of the companies that does human-data-generation-as-a-service?
Random small note—the ‘dungeon’ theme is slightly …culturally offputting? or something for me, as someone who’s never been into this kind of thing or played any of these and is therefore a bit confused about what exactly this involves, and has vague negative associations (I guess because dungeons sound unpleasant?). I wonder if something a bit blander like a story, play, or AI assistant setting could be better?
Someone who wants to claim the bounty could just buy the dataset from one of the companies that does this sort of thing, if they’re able to produce a sufficiently high-quality version, I assume? Would that be in the spirit of the bounty?
I am very excited about finding scalable ways to collect large volumes of high-quality data on weird, specific tasks. This seems very robustly useful for alignment, and not something we’re currently that good at. I’m a bit less convinced that this task itself is particularly useful.
Have you reached out to e.g. https://www.surgehq.ai/ or another one of the companies that does human-data-generation-as-a-service?
Random small note—the ‘dungeon’ theme is slightly …culturally offputting? or something for me, as someone who’s never been into this kind of thing or played any of these and is therefore a bit confused about what exactly this involves, and has vague negative associations (I guess because dungeons sound unpleasant?). I wonder if something a bit blander like a story, play, or AI assistant setting could be better?
Someone who wants to claim the bounty could just buy the dataset from one of the companies that does this sort of thing, if they’re able to produce a sufficiently high-quality version, I assume? Would that be in the spirit of the bounty?
I don’t think data companies can deliver on this complex of a task without significant oversight.