Hey @mckiev I noticed you're trading in your own market. Do you have a plan to resolve it if there's a very subjective resolution?

@benshindel Hey, Ben! Honestly, I have no idea how to resolve and would be happy to defer to someone else

so, this market is large enough for more clarification:

• what restrictions on the prompt? It surely cannot be "copy and output the following text:", maybe you have a character limit in mind?

• May the AI access the internet?

• What level of proof do you demand for the 1000+ reads? Page visits? Likes on the tweets talking about it? Just general vibe that a lot of people read it?

• Will you go look for the story, or will you burden the "yes" voters to post one in the comments?

• Is high-quality anything but meaning "got 1000+ human readers", or will you also poll, or do you have to like it yourself?

• Any way you'd resolve this N/A or percentage?

@Jono3h I'm tentatively betting no, since although I think 80% ain't that bad of a guess, I think it would be hard to find out whether it happened since most sites don't volunteer statistics like that.

Only if it became a very clear sensation would I see this market resolve to an uncontroversial "yes". But that collides with a bunch of confounders, mainly the virality of AI-generated content seems low. The appetite for AI-generated stories could be really low in a year considering the current reaction to people posting midjourney images or write blogposts that are recognisably GPT vomit.

@Jono3h Wow, these are great questions. What do you think would be a good way to resolve?

@mckiev Its hard to know for sure whether some text is LLM generated.

You could ask for prompts that generate good stories and then publish those so you can get manifold's opinion or read them yourself.

You could still have the 1000+ readers vibe by including an "or a reproducible-ish LLM story got viral" where viral is up to the reader, though you could do a percentage resolution of #tweet-likes / 10.000

I think there is a pretty big probability gap between: "there is an LLM that generates good stories when prompted well" and "I can verify that somebody made an LLM write a good story". This market's probability seems to be on the former whereas the latter I'd put at around 60%

uh and to answer my other questions:

• prompts may have 100 words

• no internet access

• make a second market which is a poll where people can put and rate prompts, if none of the prompts get an approval of 50%, resolve no, otherwise you read stories from the top 5 and rate yes if one of them is high quality

also tip me for doing your job

didn't tip me