What question will I use my free crystal on? 馃敭
Basic
39
11k
Jul 30
50%
Other
1mo
32%
How will my hypothesis that LLMs can be modeled as (something like) Solomonoff Inductors hold up?
22d
6%
Will Winston Peters hold the balance of power in the next election?
1mo
5%
Who will I marry?
1mo
2%
Who will be elected President in 2024?
1mo
1.4%
If Biden wins in 2024, will he serve a full second term?
1mo

During the Manifest 2024 opening Manifold Theatre, I won a crystal that can be redeemed for one free Crystal-tier market.

That鈥檚 a LOT of free liquidity! Don鈥檛 want that to go to waste.

So what will I put it towards?

Can be an existing question or a new one that I create.

Get 峁600 play money
Sort by:

You can sponsor the World Dog Surfing Championship market! It'll close early so you can get most of the mana back. Sponsorship means you get to claim prominent placement on the market ("Brought to you by X"), and can add some ads to the title, description, comments, and/or banner photo.

/Ziddletwix/will-carson-the-terrier-once-again

reposted

馃敭 SPEND NC's MONEY/INSPIRE A NEW MARKET 馃敭
come and make some suggestions so we can get another Crystal market!
(if you have a suggestion and don't want to pay to add it, comment or DM me and I'll add it for you)

oh it closed! 馃憖

Oops! That was just a placeholder.

Y鈥檏now, it鈥檚 only just occurred to me that I should probably make most of these markets anyway.

How will my hypothesis that LLMs can be modeled as (something like) Solomonoff Inductors hold up?

This is my PhD topic. I came up with the hypothesis that since Transformers can emulate any arbitrary Turing machine, it鈥檚 possible that they can be modeled as implementing an approximation of Solomonoff Induction, i.e. a Bayesian mixture of all possible Turing machines. I haven鈥檛 totally rejected or accepted this hypothesis (and it鈥檚 a little ill-defined anyway) but I currently think they come closer to implementing some Bayesian mixture of some other model of computation, possibly Markov chains. I have some other ideas, which could make one heck of a multi choice market!

There are a lot of smart people in AI on Manifold, and I鈥檓 sure they could give me some interesting ideas and discussions at the very least.

I鈥檇 want to give a heck of a lot of care to the resolution criteria, although I suppose I could just put a disclaimer upfront that it will resolve to my subjective opinion鈥

Ideas and impressions encouraged. If I had to choose a market idea right this second, this is what I鈥檇 go with.

I鈥檒l probably make this market anyway, and I might buy some 峁 just to make it a Premium market at least.

I will say that this idea has the important upside of it actually mattering that it's crystal. Being crystal doesn't affect a marriage market much because people won't have much insight regardless. Being crystal won't affect "Will Biden get elected" because people were going to aggressively bet on that anyways.

Who will be elected New Zealand Prime Minister in the next election?
bought 峁500 Who will be elected ... NO

I like the idea of making all the Americans on this site care about NZ politics, but it鈥檚 gonna be Christopher Luxon.

bought 峁150 Other YES

@NcyRocks Added a more interesting question

@nikki Hmm鈥 tempting鈥

If prediction markets are legalised in the USA before 2030, what will be the reason?

These questions tend to be more of a weird sort of discussion board than an actual market. I think the main consequence of a massive subsidy would be massive drama over resolution criteria

One thing that might be useful to consider: long term markets are barely helped by more subsidy. If because of interest rates it isn't worthwhile to buy YES on a 100 mana liquidity market that will definitely resolve YES in 2030, it still isn't worthwhile with 1 million mana in liquidity. the latter case just means the market is trailing the "correct probability" more. so I'd think shorter term markets are better, and even better when you can close the market before some reveal of what the resolution will be, so that you can recoup some of the subsidy mana

@Bayesian So something that will be revealed at a particular nearby point in time might work well. That鈥檚 a point in favour of an election market, although that seems kind of lame.

bought 峁60 Other YES

@NcyRocks yeah, unfortunately current manifold market mechanism seems underpowered for a lot of really interesting and important questions. but yeah those would work well.

If Biden wins in 2024, will he serve a full second term?

Hmm, my most successful market. Interesting鈥

bought 峁50 Will the WHO declare... NO

@NcyRocks Also has the benefit of N/Aing and giving you 1 million mana for free if Trump wins...

opened a 峁10,000 Coinflip market NO at 1.0% order

@SaviorofPlant I don鈥檛 think that鈥檚 how it鈥檇 work? The liquidity will be provided by Manifold, so if they haven鈥檛 changed how liquidity works then it鈥檇 go straight back to them.

At any rate, I don鈥檛 want to take advantage of Manifold鈥檚 good will to make a profit. They wanted to subsidise a Crystal market, and I intend to give them the best one I can.