Prediction Markets Panel

Transcript

[Music] [Music] we are about to begin our prediction Market I will uh hand it over to our moderator um so if everyone could quiet down while we begin prediction markets thank you all for coming uh I have a a general question that I started asking during the tea bre and that is what do you see the INF funnel and the out funnel into prediction market so what

kind of processes that could themselves be related to ethereum could lead to prediction markets be very liquid and very effective and if we manage to create liquid and effective prediction markets what kind of ethereum uh processes could benefit from that so if we could start from Martin please yeah so um as I as I mentioned in in the presentation earlier on gnosis our our goal is first to set it up uh to make it really really easy to set up a prediction Market um so let's say you are interested in one specific topic let's just have an example um decisions on Supreme Court decisions or any kind

of special interest or what's going on in the next um season of Heth of cards whatever um then it should be really easy uh as easy as setting up a WordPress block to set up a prediction market and um one addition additional mechanism that um might make make it uh easier to get started is that you can use first um tokens for those markets and um they can have a value or they they can be additionally traded in eer um but it it's for sure it's easier to to get it started uh if you have some kind of play money that only evolves over time into

something that becomes valuable because those who have a lot of tokens kind of um uh certificate that they are experts in in this field Jack um well I think one thing that uh could help sort of onboard people is like really meaningful social media media integration so um one integration that I've thought would be pretty cool is like um if you could sort of challenge people on Twitter like the I guess uh Twitter has like a change tip integration now I think they do and um you know you can imagine you know

someone is always is on Twitter saying something that's just like blatantly wrong and I mean okay so you could you could rage at that person it's not going to change anything um but you could challenge them to put their money where their mouth is which is sort of the the um catchphrase for prediction markets and say you know you have a button you can click on that says You Know Jack has thrown down the gauntlet challenges you to place a you know five Bitcoin bet that what you're saying is right you know vaccines actually cause autism and see if they take see if they take you up on it I think the sort of conflict um sort of spectacle resp that I really like like the idea Twitter integration

uh one one thing that I I think might be problematic you just brought up the example um um that um cause autism um the problem is do you get those people to rely on a specific Oracle what what would be the Oracle it if it would be the Wikipedia article then um well they can just directly I I guess they would not agree to it the problem with that example is that vaccines cause autism isn't a prediction so much as a question so then you have to formulate exactly you know what would a testable prediction off of that be yeah know that's a fair point and I mean I think that I the trick if

you if you were sort of the the Challenger would be finding you know an existing prediction Market or even creating one that matched the challenge that you wanted to place so one one sort of example that might be sort of better more effective is let's say you had two people one of whom thinks that technological unemployment is going to be a very serious problem and the other one who says well you know no humans aren't like coures we're we're not going to be replaced we're infinitely creative and we'll just keep on finding better jobs and so you actually could make you know between those two people actually could make a testable prediction something like you know labor force participation rates will continue being above half wages will continue going up for about 20 years even making

conditional on you know technological progress and so forth what's great about this process is that it actually really drives the conversation forwards and in a productive Direction when you can take a a conversation from you know this is what I think and this this is what you think to what would it take to know who was right um and that's something that really pushes an argument in a positive direction so you know I I think that kind of integrating this thing these things into communities um it's very positive for onboarding it's positive for those communities um and it's also potentially I mean depending how you design your settlement system it's also potentially something where you've got these people who have already identified with their reputations so potentially you can use that to settle the outcome

um based on those real world reputations if you if you build a system around that okay but I was wondering if I could push you in a slightly more ambitious Direction so so if you work for a sufficiently large organization you have heard the expression key performance indicator okay now suppose this organization was run on a blockchain and they have their key performance indicator and those are published as predictions on a prediction market so if the company steers left and the price of the contract tanks that's an indication that perhaps should be starting steering right instead so do you see that as a potential uh addition to the toolkit of

the governance of a company really actually I think you know in the ethereum community it seems as though lots of people are very excited about this idea of making their own ethereum startups be daos and you know if you want to create a blockchain based organization then you definitely need to have some kind of blockchainbased governance so you know having people vote as one approach um it also has weaknesses so for example one very simple one that I think people might be overlooking a bit is if you have the simple approach where you know people vote people become automatically become shareholders and shareholders vote on where the money goes then you know here's one thing that I could do let's say you know SW walk.it does their

little crowd fund does their little crowdfund and they get $80,000 $80,000 on the last in the last hour I dunk in $100,000 I get 55% and and I I make a motion to sends $180,000 back to myself and I have 55% of the vote and so I win so this is basically the reason why in the real world we have things like minority shareholder protections so basically the point of this is that governance mechanisms especially those for a system that's kind of intended to really sort of stand the stand the test of time and even without sort of relying on manual human inter intervention you know potentially ever it's figuring out

sort of effective governance mechanisms is a very a very hard and interesting problem and I could easily see projection markets playing a part in that okay anyone else on this I think uh one really nice example to to showcast uh this governance thing could be and I'm looking forward to to do it uh to to play a chess game to um the key performance indicator as you mentioned it would be will um will we win the game on the one hand the prediction market and then you could uh you could for every move have different markets and and see what's the likelihood to win the game under the condition of this move and on the other side you have a very strong player and it would be really

nice showcase if the prediction Market could beat a very strong player I think that the um you know I mean one of the examples that that's thrown around for you know prediction markets as governance is just like you know will the share price go up or down if this person is elected uh CEO and by by making it a market you know you attach a financial incentive to it and um I think it becomes more meaningful than just say a vote okay and so now let's explore the other end of the prediction now suppose you do have in place liquid well functioning efficient prediction markets how do you teach people to extract who are not necessarily directly involved in the events that are predicted to to

extract a signal out of it that can use for their activities and possibly within the blockchain itself well I think that if it if it's really well functioning um and if you've really got reliable information then actually world will spread that quite I mean and it's it's quite well known how well prediction markets have been performing for example outperforming pundits um in predicting you know who's going to win elections um so I think that once it's a it's clearly a valid source of information it's going to be no problem um getting people to actually use that information the difficult problem is making it actually a valid source information that isn't going to be gained isn't going to going to be messed with I mean I think there definitely is such a thing as you know markets sort of Market maturity and and liquidity actually jumping onto a market

because you know we see right now we already have a bunch of prediction markets on group on group gnosis Happ trading this week and I'm pleased to announce that I earned about 25 ether insider trading this week so for those who work here on Monday the story is basically there was a market maret on whether or not I'm going to wear a hat again this week um so the uh other Market there was on whether or not the price of ether would be higher on Saturday than it was on Monday and you know if markets were efficient at all we would probably see the probability of pretty efficiently kind of track you know jump down quite a quite a bit when the when the price goes down go up when

the price goes up and so forth so you know the you could EAS it's kind of easy to just as an economist to kind of think that you know if basically you know if you set the incentives they will come but at the same time it does kind of take time for that to happen and you know th those kinds of activities it's important not to forget that those activities happening actually involves you know real humans doing real things and actually sort of writing the the Bots or you know doing things or doing whatever things themselves and all of that kind of INF real infrastructural work happening so I'm actually going to latch onto the last comment so uh we seem to be in an unusual crowd here where most people seem to have a positive opinion of prediction markets

but that's by far uh not the norm and a lot of people who strongly resist to this idea are actually concerned about the possibility of somebody with Deep Pockets um moving a prediction Market in a certain direction for the purpose of taking advantage of it later so how would you guys uh uh Comfort allay the fears of such a person we start with Martin I would say that's that's difficult because uh to some degree I also have those concerns and I would well need to um really paint a bigger picture of a of of the new um economy where where

hopefully um well more people get uh directly shareholders of things they do and and then therefore make better better decisions and and therefore uh money cannot influence them so easy um but um yeah it it's it's really a bigger bigger bigger answer jump over here Ed what do you think well it's going to happen um people are going to spend money to to move markets and change information they and and you see this a lot right now um particularly in political markets because political betting example in the UK where it's all entirely legal but it's still an extremely small Market with not a lot of liquidity um compared to for example Sports and so for example um a few years

back there was a um a leadership election between uh Nick CLE and uh Chris hun and somebody was putting a lot of money on Chris hun in a way that seemed to be you know aimed at moving the price um the solution to it or or the um you know the mitigation is that hopefully if you've got enough liquidity and enough people who know what's going on they're going to pile a load of money on the opposite side and they're going to make a big profit you know um so so in in that sense I think from the point of view of people who actually um run and promote these um prediction markets it's something they'd like to see happen because all of their honest users are actually going to make money when people try to to manipulate the markets yeah me what I would say is just that um you

know what prediction markets do assuming that they're they're liquid and well functioning assumptions you said earlier uh is you know they they make it so that if you're manipulating the market you'll lose money you should lose money um and you know that now that doesn't necessarily help you in you know this hypothetical situation where you have someone with an unlimited amount of money and nothing to lose and they're just you know kind of out to burn the world um but I I think that's you know that that's a quote unquote vulnerability that you know most systems have that okay so let me dig a little bit deeper into that so imagine a situation where the prediction is about an act of sabotage where an individ

idual could actually uh by themselves or with help actually carry out the act of sabotage and cause the event to actually take place so uh what what do you guys think a responsible prediction Market should do to deal with that kind of case you could start with Edmund again yes so um I I mean as a data fee provider we've always kind of agonized about whether we should be providing um signed data about people's deaths um which obviously has a lot of potential for abuse you know people bring out the assassination contract when you when you talk about prediction markets at the same time it has all kinds of really useful interesting use cases like um you know making wheels execute quickly for example or you know uh governance

governance procedures um so we we still haven't quite made up our minds what to do ultimately about that um but I I think you know it's going to happen somebody's going to do it Mar yeah obviously also thought about is um sign messages of U people's deaths and B basically in the world of ethereum every time you are publishing a signed message you're pulling a tri a trigger of a machine where you don't know what's what's happening in this machine so you you could trigger anything with this signed message and um especially when it comes to um the death of people um it it might be or it it might uh be a solution

to say um you as an individual would have the right to make an oracle wrongly report about your your death for example to kind kind of unload the gun in in this uh so Jack um I gu my my my my my thought about that is that um you know why not take advantage of the the same mechanism that you know sort of powers prediction markets in general the idea of the wisdom uh of the crowd and you know say that you know well what does it mean for a market to be saying like unethical um and well it's really you know whether or not people think it's unethical and uh so you could sort of Leverage the moral compass of your

users and say like well they can vote on whether or not it's this is an ethical uh whether or not this is an ethical question um link yeah I mean I think as you know as I understands you know the the way augur does that is basically that when you when people actually vote on whether or not an event occurred they can vote Yes they can vote no or they can vote you know ambiguous or unethical and the theory is that you know if some question actually is an ethical in some way or if it's just sort of too plain ambiguous what like if you just throw gibberish into the question field then they'll answer that and even the person who created the question might even be penalized some amount so it's like the thing is that all of you know these

blockchain based systems they all kind of exist in this sort of cryptosphere and there's always going to be some kind of link between that in the real world and it's kind of clear that that link is inevitably going to be a collusion Vector I guess both for good and for evil okay a going in a slightly different direction now so uh imagine that you had the ability to make the life of the predictor as easy as possible so imagine the equivalent of an IDE for predictors where where you enter the market and you have a series of dashboards that you can look at that can assist you with your prediction and here there is an interesting link uh because a lot of uh governments at various

levels National Regional and even City uh are now publishing very large amounts of data sets that describe what's happening in the region or the country at various level so economic statistics demographics transportation and so on so if you had 12 months to work on such a system so H how would you set it up if we could start from Edmund again um so so so the question is that we how would we set up giving you access to all this information so that you can make a prediction yeah or which kind of data sets would you if if this was your dashboard what would you want to see or for what kind of predictions right well well it's it's going to really you know enti entirely depend on the um um on on

the the subject that thing you're trying to predict but um uh the one of the things you you want to know is what everybody else on the market is doing that that's something that um that you normally uh do know um yeah apart from that it's going to be very context defend on I think um I think one sort of importantance and con sort of convenience thing to have is just a very simple field that lets you just ask and make a prediction yourself just like fill in a text box make a prediction click yes and then you know maybe put in maybe put in some Market maker subsidy and then just see people start voting on that um I mean this is May this is maybe a somewhat optimistic answer but uh you ask for sort of the best case scenario I

so I would say you know if you had like a really good you know sort of AI machine learning uh text mining algorithm that could suggest answers to you um and you could always uh verify them yourself afterwards um I I I think that the knowing how other people are are deciding is that that that's a would also be a very strong indicator um but I you know that relies on getting something like that to work uh we we are we are sort of researching something like that uh at at augur but um it remains to be seen whether that will pan out

Martin Martin if you have anything to add okay so then I'm trying to coax a prediction out of you so um what kind so how many uh question types do you so today you know the the the most frequently cited examples of prediction Market are about political events so elections or uh Hollywood movies gross totals or um issues that have to do with popular culture in general right and so those are fairly predictable in their times uh how many different time types of predictions do you expect to see say in

the next five years time so a 10-fold increase a 100 fold increase and one millionfold increase and what would drive that growth if if possible so can we start from Jack sure um I mean I so I don't have a good sense of like the you know the the number of categories that people will be making predictions on um I think that the you know my sense is that and maybe this is a pessimistic prediction but I I I feel like the the you know the few basic things that people want to predict uh won't be that different than they are now I think I think politics Will decline I think that's sort of a you know for some reason that has become

prediction markets a uh you know the the sort of go-to topic um I think that um one thing that PE one thing that might be uh somewhat new so you mentioned like Hollywood stock exchange uh will be people um sort of betting on things like who will win American Idol um I think a lot of people would find that to be very diverting uh and you know aside from that sort of the The Usual Suspects you know you can imagine like um you know uh Financial betting I mean it's sort of like a you know if you think of like well options are kind of like betting um you know maybe Sports things like that

not you know I'm not too sure on how on exactly how long the road map until this will be but one area that I'm interested in is actually rather than sort of starting from the top and sort of thinking big of you know is making predictions on whether policy a or policy B will will be more likely to increase GDP by 5 trillion years or anything of that scale why not use them for things that are extremely microscopic so one simple example of prediction Market is you can imagine one being used to say create a voting system for a decentralized Reddit so the basic problem here is that in General on Reddit you have two types of moderation one type is community moderation up votes and down votes it's nice it's sort

of democratic and decentralized and people like that but the problem is it's slow and people don't like the fact that if someone posts a spammy Content it comment it still stays on for like at least a couple of hours before it gets kicked off the page the second kind is centralized moderation which people don't like because it's you know the big big bad guy with an Iron Fist and for everything so one sort of alternative proposal is imagine every single Reddit comment being a prediction market and people could you know up vote and down vote they could up vote and down vote themselves they could potentially you know have a machine learning bought do the up voting and down voting you know potentially any strategy goes but well the thing that they will be predicting is you know it could imagine some sort of pool of moderate of community

moderators and this could be sort of very you know very large fairly easy to join and you know people would sort of be moderating random Subs sets of of comments and this would be a prediction Market on how they would vote so it's sort of kind of like I think Robin Hansen came up with what he called predict cracy which is like using prediction markets on voting so this would be kind of similar so the benefit here is that you know people could make sort of down votes and potentially push content down very quickly but you know by doing that they're putting their money where their mouth is and if they're abusing the system then theoretically someone else could sort of bet against them and steal their money so you know that's just one very you know very simple sort of you know completely out there example of how

prediction markets could be applied on like Sub sub dollar scales so again let me build on that so I find it uh interesting and remarkable how still web oriented much of the discussion is but suppose you want to know um how many new stores a large company like Walmart will open in the United States next year uh you really want to have the people who are driving by the construction sites of the Walmart site to tell you if the construction is proceeding on schedule so what kind of mechanism do you think could be put into place to have those very microl local observations feed into a macro process like how well Walmart is going to do next year make this sort of haakan ideal

of markets as sort of decentralized information aggregators actually really work well um I think I mean the problem seems to be a user interface Problem by and large so you know the solutions are going to be sort of things that you know economists might not find to sexy but which app app developers perfectly well understand like you know making sort of nice and easy to use apps that people can sort of you know see a bunch of prediction markets potentially even you know sort of use machine learning to sort of suggest which ones you should look at and just make it very easy for people to bet on them one thing that that's important in my mind is that you can have very very

different views on the same data so um so one reason to to place a prediction on a prediction Market could be hedging against the risk and basically buying an insurance um but that would obviously be a completely different uh user interface and and look and feel uh but still it could it could access or it could um be on the same market and then could be on the other side or on another side a market that's more spec speculative on on on those um and I think those yeah different um use cases might also help um to yeah to to get the ecosystem right

for a prediction Market because in the end it's a it's a zero sum game so in the end if someone wins money and and those who contribute the information they they should win money then someone else on average needs to lose money and that could be someone who for example wants to hatch against the risk right so that's like the famous nor trade uh uh theor that says that if somebody's taking that side of the bat who am I to think that I know more than they do okay so again how sometimes they might so how how do you break that uh uh that problem well I was just going to say like wholeheartedly agree with the the I mean really what both of you guys have said

just that the the user interface is is very important and and especially with your point that you know really you there there is not necessarily a one siiz fits-all solution for for user interfaces in general but for prediction Market specifically like you can imagine that um I mean you can break it down sort of as granularly as you as you want but the you know it seems like well you'd have like the sort of uh Traders would want one kind of interface they want really all the details all the data presented to them you if you want maybe people who are it seems like people who are betting on say sports would would expect a different interface right they they they'd expect like odds instead of

probabilities um and well and maybe the most important one is just sort of a newbie interface one that doesn't require you to sift through mountains of data or understand how odds work or really anything like that you know so just an app where you can click like on you know where you drive by the Walmart and is this going well yes or no that sort of thing okay so uh different question so imagine a situation not too far I don't think where a very Advanced artificial intelligence gets access to one of these markets and systematically uh ends up uh uh outbidding everyone so what would happen to a market in those in those

circumstances that's it sounds like you know at the very start people would lose and then in the second phase people would kind of realize that the AI is probably is smarter than them and so there's no point in disagreeing with any of the probabilities and then of course the third phase would be that the AI realizes that aside from this crazy ethereum thing there's this thing called Wall Street where the AI can make millions of times more money and so the the the AI would vacate the market and move to better things and leave the humans uh pick up the scraps that would be nice but in the meantime you'd have like perfectly efficient prediction markets that always told you the right

answer a good thing oh I'm up okay so thank you all for uh being on the panel and I HED everybody else had fun thank you absolutely [Music]