This market will resolve yes if Manifold has a feature that allows commonly used keywords to have a definition Manifold. The keywords must be able to be used across markets and creators.
What I see today is:
Manifold has a commonly used keywords such as "State of the Art Large Language Model" or "Superintelligent Artificial Intelligence", where most market creators are meaning roughly the same thing but there ends up being many different definitions due to market creators having to redefine it in each new market.
What I expect this feature to be like at a minimum:
Manifold creates functionality for a dictionary/wiki type page where someone can define "State of the Art Large Language Model (SOTA LLM)" or "Superintelligent Artifical Intelligence (SAI)", with a relevant definition and the benchmarks/inclusion criteria that would be required for something to be considered a SOTA LLM or SAI.
When creating a market, creators are suggested defintions for keywords they are using (e.g. they type "This market will resolve Yes if a new SOTA LLM is created by 2024", and then manifold/clippy says "Would you like to use the definition for SOTA LLM found here?). Then, the resolution criteria of that market include the commonly understood definition.
I think this would be very useful as:
It would make it easier to create high quality markets!
It would make it more valuable/easier to compare markets on what are currently similar (but frustratingly, not the same) topics
If it becomes a part of search/groups, it would reduce the incentive for making low quality markets
Please feel free to make suggestions or ask comments clarifying the resolution criteria in the comments.
Some (possibly overly complicated and definitely not fully thought out) ideas for how I think this could be a commonly used, positive sum feature are (not required for this to resolve yes):
Makes definitions a new type of "group", and
weights markets using the same definitions in their reccomendation algorithm so they incentivize accurately tagging markets and using definitions
allow advertising of questions as a whole using a given definition/group
Creation of definitions could be incentivised by a bounty or usage fee (maybe start with a quest and/or reward for creating them too?)
Definitions could be minted, e.g. when someone wants to create a definition they create X "instances" of it (paid to manifold at a rate of Y per instance), then they establish a baseline price they are willing to sell usage of it at, then people pay said fee to use it (maybe include a price curve resonding to supply/demand, and an automatic way to mint new instances if price goes above X+Z?)
Definitions/definition content could be decided on with futarchy; each time someone creates or updates a definition, a derivative market is created asking something like "If the definition for "SOTA LLM" is created/updated, will the market clearing price for it be higher or lower than X (where X is the current price)."
Hoverable tooltips like wikipedia!
initially, subsidize the usage of definitions with a quest/reward for making markets with them (or updating your existing markets to use them).
PS: If someone makes this (by market close), it gets integrated into Manifold, and it works in a way I would expect to be beneficial (on release), I will pay $100 (including if it is done by a manifold engineer). If someone includes beautiful (as decided by me) market mechanics/futarchy into it, I will pay an additional $100. I reserve the right to pay out portions of this if I believe it is only somewhat implemented.
Related questions
๐ Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | แน106 | |
| 2 | แน31 | |
| 3 | แน17 | |
| 4 | แน11 | |
| 5 | แน10 |
I just use mc unlinked for this and put all data in description
I think this kind of thing would be really cool! I think a MVP might be to use an existing wiki solution, and then people can link to the definitions. Unfortunately, I don't know of any good wiki software atm; Notion is pretty good but not super well-suited to public-facing hosting rather than for teams.
(I've long thought Manifold should consider in-housing a wiki, eg by extending something like Posts, but then I also generally believe Manifold should support every possible architecture under the sun)
I'm not sure that the specific use case of wikitionary-like "market definitions" is very compelling as it doesn't feel like being unclear on "SOTA LLM" is a very common problem. But more broadly, wikis that collect eg the most important questions regarding GPT-5, or helps someone explore the space of Ukraine forecasts, seem really really nice.
@Austin extending the posts section is what I had in mind, although using external software for the wiki features would definitely count. I personally thought using external software would be more work than extending posts - they currently would count assuming you suggested usage of them when writing keywords in markets and added an edit functionality (which while not required for the resolution of this market and Bounty A, imho should totally be done through futarchy).
The reasons I'm quite long on the definitional part of this idea is that I think it will increase the median markets quality a decent bit through making it harder to have poorly defined resolution criteria, as well as making it easier to compare unintentionally different markets (which I find makes low quality sorta identical markets useful, since you can try figuring out the distribution of outcomes in a given area).
There's sooo many low quality markets out there. They lack any resolution criteria other than the title (and maybe if we're lucky a non-value add restatement of the title in the description), so it's not worth it to bid on them as is (and it gets tiring to constantly comment "OK, but what do you mean"). I've ended up just blocking a bunch of different groups/users I find correlate with low quality markets, but this is reduces the value of the site for me as a lot of them could be decent (if they just bothered to define their terminology ๐ ).
Further, what kicked this idea off in my head was I wanted to make markets asking roughly "Will a SOTA LLM originate in China by 202_" which I think is a high value question, but to make decent resolution criteria have to research benchmarks that show SOTA performance for LLM's and find ones that also include international research teams. I'm not an SME, so I don't know those off the top of my head. I'm sure someones done this on Manifold before, but it's not easily accessible, and even if I do find their relevant description and copy/paste it, we have standard DRY related issues. If I don't find it, I'm sure I'll end up making a slightly different definition than them, but then we lose the ability to compare our markets probabilities to eachother as well (since our independent variables aren't as isolated as they should be). (I /may/ also be unusually pedantic and that is the root cause of my frustration).
On collecting/ranking high vs low quality questions, I definitely think there is value there (although I don't think my proposition more than incidentally does so). Maybe have definition/group based leagues (for each definition/group)?
@Conflux but how many feature requests offer a truly gamechanging 200 dollars bounty if a manifold dev feels bored/doesn't have as good of an idea at their next hackathon? ๐