Okay, we're going to get started. Thank you everyone for coming out.

I think I know most of you. For those I don't know, hi, I'm Hannah. I'm one of the organizers.

We also have Ian here, and Chris, who I think has exited the room, but he's around somewhere.

So we're going to talk about prediction markets today with Paul.

But first, last chance to participate in the raffle if you want to win any of the cool stuff back on the table over there.

The Don't Trust, Verify, Hats, Cold Cards, Open Dimes, that good stuff.

You can go to that website right there, bob.shy.org, slash donate, and buy some raffle tickets.

We're going to do that after the talk.

So we are recording this, so if you don't want to be on video, just stay back there.

I don't think we're going to record Q&A. Are we, Ian, we're going to stop it there?

Yeah, okay. So the Q&A, when we get to the Q&A section, you will not be recorded.

Pizza will be coming later. We got some really generous donations today, so thank you to people who donated.

So we have the budget for pizza. Yay.

So that will be here soon. Please leave it there until we get to the Q&A, and then we'll all get up and chat and eat pizza.

We are going to mosey on down to the bar after this if anyone would like to join us.

And we do do lightning talks before the talk.

So that's – anyone can come up and tell us, you know, about some project you're working on,

whatever the case may be, for two minutes.

So do we have any takers today? Does anyone want to come up and tell us about something?

No? Shy crowd? Oh.

I just really want to quickly announce the conference is coming up.

Okay, we'll make that –

In case anybody doesn't know.

Do it properly.

I'm shy.

Keep it brief. Come up here.

Yeah, we'll be very brief.

So for anybody who doesn't know, we have a large conference coming to Chicago called White's Blockchain.

It's the second year.

It's going to be at Benju 610 on September 30th and October 1st.

That's Monday and Tuesday.

There's three stages over two days, 200 speakers, Fidelity, TD Ameritrade, Deloitte are among the sponsors.

And we have speakers from – everything from researchers with Monero and Enigma, SDGrad,

to the corporate side where we have, like, Samsung, SAP.

I think it's falling asleep.

And a bunch of corporate stuff, and then there's a track for the digital assets and markets as well,

obviously with TD Ameritrade and Fidelity.

And if anybody wants a discount, there's a code, BOB100, all caps, B-O-B-100, for $100 off tickets.

So I just want to make sure everybody knows it's happening.

All right. Thank you.

Anyone else want to come up and tell us anything?

No? All right. Cool.

On to the main event then.

So I think probably most of you are here because you're already familiar with Paul's work, but in case you're not,

we'll run through it briefly.

So you've been interested in market governance and prediction markets your entire life.

That's true. Yes.

Yeah, if you can stop it from falling asleep, that would be great.

That would add a lot of suspense for me.

I think.

That got lots of attention and lots of interest, and he left his statistician job at Yale to go build that.

That's true.

And then in 2016 started working on Drivechain, which is another really cool project.

We're not talking about that one so much today.

We're talking more about prediction markets, and you also do a lot of research,

which you publish on your website, Truthcoin.info.

That's it.

And he's going to be talking to us today about prediction markets.

We're going to do a bit of Q&A during the talk, and we'll also have a Q&A session afterwards.

You have to unlock it.

Unlock what?

Oh, yeah.

Slight technical difficulties.

Standby.

This is a very secure event.

All the projectors and the Wi-Fi, everything is extremely secure here.

Yes.

There we go.

Okay, good.

That would have added a lot of suspense to the talk if I had to move the slide once every one minute

or whatever that was before it shuts off the thing.

That would have been something.

Welcome, Paul.

Okay.

Thank you.

Thank you very much.

Thanks, guys, for inviting me.

Chicago is a super cool place.

Yeah, so we already kind of went through this, so I guess I'm not sure if I really should.

I'm from Summit, Connecticut, and I have a background in economics and some math and statistics.

I did work for the Yale Department of Economics.

The guy I worked for, Bill Nordhaus, won the Nobel Prize.

He was the most recent winner of the Nobel Prize, which is kind of cool.

So I feel like I should take credit for it by mentioning it.

And then I have a bunch of Bitcoin.

You know, I speak at Bitcoin conferences, especially the scaling conferences

and some other conferences, and that's really great.

But this is a talk about – oh, this is not sure.

Can we zoom maybe?

I don't think we are getting it 100%.

It's not like full screen.

Do you know what I mean?

Like it's kind of –

Oh, there we go.

Oh, yeah.

There we go.

Oh, they made it much worse.

All right.

Let's see.

Yeah.

What is it?

Yeah.

I don't see that.

Oh, there we go.

Okay.

Nice.

This is a very – okay.

No problem.

Okay.

So this is a – the topic of this presentation is an idea for getting the public sector

– and it's this scary red democracy blob –

to substantially inherit some of the reasonableness of the private sector.

So it's not about replacement, which most crypto cypherpunk stuff is.

It is not about replacing the public sector with the private sector somehow.

So it's about importing some features.

And here's the talk in one slide.

The problem is that politicians install bad policies into society and they get away with it.

And the solution is to make it easy to detect the imperfect politicians and get rid of them.

And specifically, we're going to create some blockchain assets whose ultimate valuation reflects a politician's competence.

So citizens are just going to glance at some prices right before they vote,

and they're going to pick the – they're going to vote for the person who has better numbers.

So people might have this information, for example, just before they vote.

And for every person running, those are the columns.

They're going to give forecasts about how life could be if the person is elected.

So you just compare.

The first row gets how much the government will cost.

And most people know that they're paying for that.

And the second row is how economically prosperous the citizens are.

The third row can cover all kinds of things like invasion, health care, mental health, gun safety, whatever.

That's kind of a macabre one, but this is just an example.

A person will just – a voter will just look at the two columns and they'll just say,

well, this is a better number, cheaper government, and I might earn more or someone will earn more.

And these numbers are kind of like the same.

I mean, this one's better, but, you know, they're similar.

So the talk is where do these numbers come from and why is anyone going to take them seriously?

Because it's easy to just put up a table here, but this is the –

so the talk is going to give – there's going to be some background knowledge,

and then I'm going to talk about why I think elections don't – you know, why they have trouble working, let's just say.

And then I'm going to talk about event derivatives, and then I'm going to combine all this.

And then the last point, we'll just ramble on and on for as long as it takes,

but it will be about how to get this idea to be persuasive enough to actually make a difference.

So – but the meat of it will be the first four points.

So the background knowledge that you need to have is that this idea is very old.

I even have a slide later.

This goes back – the general version of this idea goes back almost 1,000 years,

and there's even a quote about betting on who will be elected pope.

And this idea having been described as very old and widespread as of like the year 1507 or something.

But even this specific idea of futarchy for democracy is also pretty old,

and it's actually a favorite of a – well, okay, let me just stick to my notes here.

This is why I always do this.

Anyway, so the second thing is that I basically already implemented the idea,

and the third thing is that there's no kind of weird scam ICO or any other kind of weird scamming here.

So I'm going to drop some names.

This idea is a favorite of Hal Finney's.

Hal Finney wrote this post.

A lot of people think that Hal Finney was Satoshi Nakamoto.

He received the first Bitcoin transaction.

The idea is also a favorite of Robin Hanson's, who was endorsed by Ralph Merkle.

Ralph Merkle built most of the technology that ended up in Bitcoin.

This is a paper that was put out in order to convince the government to stop interfering with prediction markets

or event derivatives, as I prefer to call them.

It's signed by a ton of people, and a lot of them have won the Nobel Prize or will win the Nobel Prize,

or they're just these huge names.

So Hal Varians, chief economist at Google, Tom Schelling, Ken Arrow, these people all won the Nobel Prize.

This is just like a big – it's a big widespread list of extremely elite people.

Put out a little two-page paper to try and get some types of gambling less regulated,

but they didn't succeed, hence the blockchain strategy for this project.

So this talk assumes that someone will solve the so-called blockchain oracle problem in order to make these assets.

And I think that I've solved it, so there's nothing for you to worry about.

I wrote this paper about it a long time ago.

And here are some reviews of the paper, because going into that paper, that would be a lot of work

and it would make for a terrible talk, but you just have to take some of these people's word

and just kind of roll with it as being whether or not it's still worth your attention to pay attention in this talk.

But yeah, you can see there's Roger Ver, but there's also Adam Back, and then there's Peter Todd and Andrew Polster.

I think Andrew Polster is probably the brightest mind in the space, his critique of design-wise.

And so the paper spawned many, I would say all of the projects in this area,

some of which I'm more proud of than others.

But this one I like is the one at the top, which is that I oversaw its development personally.

It's a fork of Bitcoin Core, and it was assembled by two Bitcoin Core developers over a period of about 24 months.

And I run the project website where I put all the content.

Like today's presentation, we'll eventually end up there.

And here are some screenshots.

I just provide these to give you an idea that this is all test, fake information,

but there's still a lot of Vaporware projects in the space,

and so I just thought it would be cool to have some screenshots for people to look at.

This information here, this price data, will be ultimately where those numbers from the table come from.

Here's some cool screenshots.

So theoretically the project could be live with real Bitcoin by the end of this year,

but that's how software works.

Probably it'll take another 100 years in reality, but it's almost there.

But that's what everyone will say when they work on software.

And I finally have to mention that there's no ICO or utility token.

There is a second token, which often leads to misunderstandings,

but the second token is more analogous to mining.

So in Bitcoin, if you transact, you don't need to mine.

In this project, the users don't need to know anything about the second token,

and they don't need to own the second token.

So only if you want to go into that line of work would you care about it.

So the customers use Bitcoin only to create and trade in all the markets.

It's a Bitcoin-only project.

Okay, so that was some background information.

So why is the public sector so disappointing?

It's not like everyone hates it. I have a congressional approval slide later, I think.

But it's not just like libertarians are weird cypherpunks.

This is a mainstream culture that the public sector is terrible.

And it's really funny. You've got to watch.

YouTube has Best of Mayor Quimby.

It's really so funny if you are a libertarian person,

and you just watch the things he says,

and he blames problems on immigrants for no reason.

It's really funny. And people just eat it up.

And then, of course, the famous South Park episode,

the giant douche versus the turd sandwich.

I couldn't put that up.

But it's just so funny.

And when Eric Carpman rolls into the scene,

and he's trying to get people to vote for the candidate he likes,

and he just says,

this is the most important election of our lives.

And then you're just like, do people fall for that?

But then in the real world, people always fall for it.

It's very funny.

So you have to have a theory.

In order to improve something, you have to say, why is it broken?

So my theory is really about feedback.

So this is a basic feedback loop in the free market.

You have a terrible product being sold.

Customers hate it, and they stop buying it,

which punishes the merchant.

You've got to make or sell something that carry goods that people want.

So a free market doesn't mean instant utopia, right?

But it does mean that things improve.

You have some recourse if stuff isn't good.

Things improve basically as fast as they possibly can.

So the public sector also has feedback, which is called voting.

And when the government is too unsatisfying in any way,

it can be dismissed with a majority vote.

Now, it's a little more complicated than that,

but it's not important to really go into the details of it.

This is basically the gear of the machine.

So you're firing people with this secret ballot.

So the free market feedback differs from electoral feedback.

We know that's the case because one of them works pretty well

and the other one doesn't really work that well.

Frequency is one difference.

The market feedback fires every time a merchant loses a sale.

That's millions of times per day.

But the electoral feedback fires once every four or two or six years,

however often the election is.

But I think that difference is very small.

The real difference is this idea that I'm going to call research

because I couldn't think of another name for it.

And there's more feedback than you originally thought

because when a disappointing sale is completed,

the consumer suffers and the desire to avoid suffering

stimulates the consumer's search for alternatives.

And I call that search effort research, for lack of a better word.

And the interesting thing is that in both sectors, public and private,

you have to do your own research and you have to pay for it.

Research consumes your scarce resources, especially your time,

but not limited to your time.

And the important thing is you research a lot.

So if you're buying a house, you research a lot.

But if you're just picking out a hotel room,

you still do a little bit of research, but you don't do that much

because it doesn't matter as much.

You'll be out of the hotel room soon.

But if you want to do some cosmetic surgery or something,

you would do a ton of research.

But if you're just buying a new pair of shoes, less research.

The problem with elections is that your single vote will almost never affect the outcome.

So people pretend that their vote matters in order to feel important,

but really they know it doesn't matter.

So if you master the issues and the candidates, you just wasted a bunch of your time.

It won't actually change your life in the slightest,

so there's no reason to really look into it.

This is obviously a big oversimplification,

but I think it is the main thrust of what's going on.

Voters usually admit that they don't know what they're doing,

and 71% of people can't even name their congressional representative.

You only have one of those, so it's even easier than remembering your senator's name.

There's two of those.

But if you can't remember the name,

how are you supposed to compare that person's performance with the rival?

It's very funny.

A lot of people, they do these gotcha questions where they ask people,

like, who was Hillary Clinton's running mate?

And people, like two or three weeks after the election,

people just forget.

People just don't even remember.

But I don't blame them because it's rational, as I've just indicated.

There's no reason to look into it at all.

It's just a waste of your time.

So on election day, if you go and get breakfast somewhere,

you're going to weigh all the costs and benefits of different breakfast foods,

but when you go in to vote,

you won't really know anything about the costs or the benefits.

You just haven't looked into it at all.

And I know people who have admitted they go in there and they're like,

this person had a nice name or whatever.

They can't remember a single thing about it.

Or they have no idea.

You go in to vote for president,

but then you have to vote for city council or something.

And you're just like, I don't know who these people are.

People make a zigzag pattern or something.

So that's what we're dealing with here.

It's just like controlling society.

But an implication of this theory that the research is the issue

is that the strategy of educating people in order to make them libertarians,

which is very popular, or make them experts in economics,

which is also, you know, that strategy won't work

because it will never solve the problem.

It's noble and worth doing in the same way that chemistry and music are worth doing

and like bird watching is a real skill

and mastering the skill of bird watching can be fun.

But ultimately, if persuasion were effective,

then the elections themselves wouldn't be so problematic.

The lay person is never going to be as smart as a specialist by definition.

The specialist has looked into it with their spare time.

So, you know, if you think that Friedrich Hayek is right

and that knowledge is diffused among lots of different people,

then it's actually going to be tough for you to have the experts be as smart

as the 51% of lay people who will be voting.

And teaching them is just another way of saying that you can't solve the problem

because the research incentives are the whole problem in the first place, as I've argued.

So that's me complaining about what I think the problem is,

but now this is where the solution starts to take place.

And as I've said, the solution is to have people show up on election day

and open their cell phones and check some prices.

And it should be as easy as checking an NFL score or the S&P 500

and then they just vote for the candidate with better prices.

But the question is, prices of what?

And this is an example from InTrade.

So we have the InTrade example.

One of the first of many, maybe.

This is a graph of prices of something called a prediction market or an event derivative.

And it's an asset that pays you money if something happens.

If it happens, you get paid.

If it doesn't happen, you don't get paid.

And so this was created in 2011.

It ran from 2011 to the end of 2012.

The bet would have been worth $1 if the year 2012 had been the warmest year on record,

as reported by these NASA satellites that measure surface temperature.

And they put it on a website where anyone can look at it.

If that doesn't happen, the asset's worth zero.

If it does happen, the asset's worth $1.

This is the price in cents on the right.

And the asset traded at 40 cents in January 2011, over here, 40 cents.

But then as 2012 went on, this is January 2012,

it became clear that 2012 would not be the warmest year on record.

And so the price collapsed.

But 2014 was the warmest year on record for those of you who are interested.

So this topic often shows up, betting on elections,

in a way that is slightly interesting but annoying for me

because people are interested in the least subversive aspect of it,

which is that people like to do this type of betting, betting on who will win.

Oh, excuse me, I had a pretty good note and I'm going to go back to it.

The big difference between this and betting,

with betting, the odds are fixed, usually at 50-50.

But that's not very interesting.

With these assets, the market price is going to co-vary with everyone's behavior.

And so that is why it can change and reflect the objective likelihood of the event in question.

Anyway, as I was saying, people like to use election betting in this way,

and this is the type that is popular for many hundreds of years.

There are assets that pay out of the candidate.

These assets pay out if the candidate wins the 2016 Republican presidential nomination.

After the Iowa Stroup poll, Trump's price plummeted and Rubio's did very well.

But eventually, when Trump won, whatever, South Carolina,

yeah, it surges up to, and this is about 80%,

so you can see that various events happen and make certain candidates more or less likely to win.

But this is not what I suggest that we do, and I think I put, yeah, I put a big N, no,

because this is very boring.

And I'm going to describe what we should do instead.

This certainly makes election years more fun, but it only tells us who is likely to win,

and that's not what this idea is, the Hal Finney, Robin Hansen idea, my idea,

is to change the outcome.

We want to change who wins.

We want the bad people to lose, and we want the good people to win.

And so in order to explain that, I have to explain just a tiny bit of statistics,

and this is nine slides.

The first two are super easy, and four of them are really one slide that is just broken into four pieces.

Anyway, so on the left, I have a series of coin flip events.

In this table, we are describing the behavior of the 303rd coin flip event,

which we can't see because it's in the future, or it's unknown for whatever reason.

We don't know what will happen in the flip, but we know that we can fill it in.

50-50, it's so easy.

It's the easiest example ever.

That's why everyone uses it first.

This is the same thing, but with a dice roll instead.

There's some list of dice rolls somewhere in the universe,

and we don't know what the outcome will be,

but we know that there are six, and not one is more likely than the other,

so each has one sixth probability.

So those are the first two that are super easy.

This one starts to get a little difficult.

This example is the previous two examples smushed together.

So we have the coin flipping green along the left vertical edge.

We have the dice roll across the top horizontal edge,

and I have some probabilities circled in red.

The probability that two will be rolled is one sixth,

and the probability that our tail will be flipped is still one half.

It's all the way over there.

Those are called marginal probabilities because they're written in the margin, so to speak,

and the probability that both of those events will happen is one twelfth,

so probabilities multiply together.

Now we'll talk about relationships among events.

I'm going to draw a contrast between these two tables here. The first table has coin flip 305 and coin flip 304.

They're both about unknown future coin flips, but in the second, something very unrealistic has happened,

something very bizarre, which is that someone has plotted coin flip 304 against itself.

It's only useful for teaching purposes. This would never happen in the real world.

But if it did, what probabilities would we write in the table? We know what we would write,

and it's not 25, 25, 25, 25. It's 50 along the diagonal and zero because 304 is either going to be heads or it's going to be tails.

And then I have the example again with the dice.

Say that some crazy person again is plotting the same dice roll against itself.

Well, if the dice is either going to be four or it's not,

so the probabilities are one sixth across the diagonal and zero.

So what I'm getting at is that there are a lot of values that are logically impossible if the events are related.

So the more related the events are, the more they clump up along this line.

So relatedness equals clumping is the point.

So all we do is measure these two things at once, and we look for the clumping.

So you build assets that only pay out on the entangled information.

So you say instead of just will candidate X be elected,

and separately, instead of just will American voters earn higher incomes,

you build something that only pays out if they both happen, if the candidate is elected and people win.

So it's zero otherwise.

It's very likely to be zero now because there's three ways of it being zero and only one way of it being worth a dollar.

So of course, as you divide the probability space, they all get smaller,

but you can buy them in pairs or sets and things like that.

And so what you ultimately do is you make these four different markets,

and you divide them by the probability space.

And so what you ultimately do is you make these four different markets for each,

and you just look for the clumping.

If it clumps, then they're related.

So then it would be candidate X is causing American incomes to go up.

So that is how you get to...

Anyway, this is the synthesis.

I think I'll skip this part.

So this is where these numbers come from.

Anyway, the laypeople never have to see anything complex.

They just have to see this table.

And that is how you get these numbers to show up.

Maybe I should actually emphasize this.

You have these, and you just do a little adding and dividing,

and you get whatever the situation is.

You can just reduce it down to this one conditional likelihood.

So here it's like Elon Musk has basically no chance of being elected.

His odds of winning are less than 1%.

But if he is elected, the economy is not going to do poorly.

All of his, 100% of his .009 points are in the yes.

The value that he's got, all of them are in the yes.

But regular voters wouldn't need to do that or do any arithmetic.

You could just show them this table.

And this is where all these numbers come from.

Now there's a lot to say about why would anyone ever care about this.

So the plan collapses if no one cares enough to trade in the markets

or look at them on election day,

something I call no trade and no attention.

So no trade, how do we get knowledgeable people to trade?

It's actually a pretty severe problem.

There's more in trade.

We've got three more in trade screenshots.

So this is what happened on in trade.

There were low trading volumes.

As someone who loved the site, it was my favorite site really on the internet at the time,

more than half the markets had zero trading as far as I can remember.

This was pitifully low.

Here's just a few that I could find on the internet.

And the volume is these bars and I deliberately changed the saturation

because it was almost invisible before.

But maybe you can still see some of them a little bit, these lines here.

And this is the volume scale.

It's quite pitiful.

In trade is so long ago that I couldn't remember

if this is number of contracts or notional value.

Since the price on in trade always ranged from $0 to $10,

you can only be off by half an order of magnitude.

And the price here is $150.

Anyway, the point is that this is on its best day,

something about whether or not Obamacare is declared unconstitutional.

That was like $1,000 in volume in a day

for something that affected obviously trillions of dollars in healthcare expenditure.

And then this is something about Hamas and Israel

that apparently no one cared about at all.

It had like $40 of trading in a day.

And here's something important.

The U.S. entering a recession in 2012,

but no one cared, $60 per day on its absolute best,

assuming that it's notional value instead of number of contracts.

These are the only markets that I was popular enough for someone

to screenshot them and write like a blog post about them

so I could even find them.

This is an issue.

It's greatly, I think I'll come back to this.

It's greatly exacerbated when you split the market into more states.

So you need at least four buyer-seller pairs of people

to get a full table of prices.

And the problem is, oh, here we go.

And yeah, and I want there to be many states

because then we can get many conditional probabilities as I explained.

So we want to like make this thing really big.

But this is something with 24 states,

and you'd really need people to be buying and selling each of those.

So this is a big problem.

Yes.

The problem gets much worse

because prediction market transactions have no consumption element.

They're 100% synthetic.

They're like artificial markets.

So in a prediction market, you're only betting

when you think everyone else has it wrong.

You're a speculator.

So you need to believe that you're smarter than all the other traders.

In contrast, when you buy milk over here,

it's just because you want to buy milk.

You don't like go look around the supermarket and think,

aha, all these suckers think they know milk pricing,

but they don't know what they're talking about.

This milk's a steal at $2.99.

They're going to regret this later or something or whatever.

You just want to buy the milk because you want to consume it.

And it's actually, I think it's the same

if one person invests in the stock market for retirement.

Obviously, you prefer to get the best price you possibly can,

just like you prefer to get the best price for milk.

But really, you just want your portfolio to have a certain risk-reward thing.

So you have a reason for buying them.

And prediction markets do have some consumption elements.

You can use them for insurance or for recreational gambling.

If you use them just to have fun,

then you're just having the fun that you wanted.

And you can use them for insurance.

I'm going to bet that Trump wins the next election

because I don't know what I'm going to do with my life if he wins,

so I want at least to get a ton of cash if he does win.

And in that case, it's more like a consumption thing,

but in general, it's very synthetic.

It's this very kind of artificial thing that's happening.

And so this all makes it much worse.

But fortunately, there's a way of solving that problem completely,

and it's actually the same solution as the second problem.

So I'm just going to ramble about this second thing and then get to the solution.

But I think this type of thing that does interest me tremendously,

this is the second problem of no attention.

So how do you get people to care on election day?

Why would it actually make any difference

when lots of other people have tried to make a difference in the past

and try every four years using everything they can to push for their side,

and they don't make a difference?

Well, we have to consider some facts,

which is that election outcome is very sensitive to voter turnout,

and voter turnout is very sensitive to this lack of knowledge or this frustration.

And prediction markets provide knowledge which can influence outcomes.

So I have some slides about this.

One is that people will be on their phones because they'll be in line.

So this is the first Google result for election line,

and there's two people who are on their phones in the screenshot.

So we've got them on their phones. That's the easiest part.

But the other thing I wanted to list of things that I want to mention

is that the victory margins are actually very narrow.

This is the presidential popular vote.

It's actually pretty narrow,

and you can see the red and blue bars are about the same length.

And voter turnout is very low.

You can see how low it is here.

For the midterm elections, which are the congresspeople who actually make all the laws,

Congress is more important than the president for, like, good policy, I think, in my opinion.

But you see for president it's like 60%,

and for the midterm congressional elections it's like 40%.

It's very low.

If all the people who didn't vote formed their own political party,

they could beat the other two parties combined.

So it's actually quite low.

And you may think that maybe it's low just because everyone's okay with what's happening.

Everyone's just like, everything's going great, so it doesn't matter because I won't vote.

But actually, everyone really hates Congress particularly.

This is actually the low, the one with the low turnout.

And you can see it's in our modern era.

I actually do think it was below 10% for at least,

I remember Jon Stewart making a joke about it before he went off the air.

Someone, he had Nancy Pelosi on or something,

and then right before I cut the commercial he said,

good luck getting that approval rating back up to 9% or something,

and then she tried to say something and it just cut away.

That's funny. That's funny. He's a funny guy.

Anyway, so why, when people are surveyed anyway, why don't they vote?

They say, well, you know, they hate both the candidates

and they don't know which candidate is worse.

I mean, that's the reasonable position to take.

And they just think, they don't know what to do.

But that's exactly where this project could help.

So at this point, the critical, I have to say, the FUD agent here,

I have to say, ignore all prediction markets because they're stupid,

or prediction markets are great, sure,

but this one in particular about my favorite candidate, whatever,

is one you should ignore.

So I'm going to do these one after another.

The first, though, is going to be tough because it actually has

a strong track record of accuracy.

Since they do work, people will learn over time how accurate they are.

The other thing that interests me is that it's actually a very old idea.

I have this and I have this other one about Google Trends for InTrade

and some CNBC appearances.

But in general, this idea of betting on elections and obsessing about it

is actually extremely old.

And I have this quote here about whatever,

wagering on selection for offices in the Catholic Church.

Odds on papal succession appear as early as year 1503

when such wagering was already considered an old practice.

There's a bunch from the U.S. I have two down here.

Well, I'll just read this last one.

This is from Andrew Carnegie in 1904.

From what I see of the betting, I do not think that Mr. Roosevelt will need my vote.

I am sure of his election.

Andrew Carnegie in 1904.

The New York Times used to publish the prices.

So actually, I think it is.

People will warm up to it.

I think they already like it, actually.

The strategy is to ramp it up over time by betting on things

that people already love betting on,

like Super Bowl, March Madness, Oscars, Tonys, whatever.

So you can get people familiar with the institution

in a context that is not as subversive.

And then they'll eventually have to say that prediction markets are great.

So we're still...

Oh, I skipped two slides, but that's okay.

What I wanted to finish my thought is we'll still have this second problem

where someone will say, sure, prediction markets are great,

but this is one you should ignore.

So let me get back to that, though.

I do want to mention that eventually almost every source of information

will be biased or captured by people who want to use it.

It's like how they get Bruce Springsteen to endorse someone,

which is like, I love Bruce Springsteen, but what difference does it make?

He endorses a Democratic candidate.

What does it mean?

But the point is that...

So one example is debates.

Everyone, when a Democrat watches a debate,

they're certain that the Democratic candidate won.

And when a Republican watches a debate,

they're certain that the Republican candidate won.

And that's because this source of information about the candidates

requires some kind of interpretation.

But in these prediction markets, there's only one interpretation.

It's just what numbers the people are probably going to hit.

And it's the same interpretation for everyone.

So people have to fold and say that these markets are unhealthy

or manipulated or something, which is what I'll get to.

Another thing is that it's tiresome to listen to all these politicians speak

and all their supporters speak about things,

because everyone knows that it's all nonsense,

but we sort of put up with it anyway.

Like people will say, oh, my favorite candidate is Andrew Yang,

but he came out for a higher minimum wage,

even though I'm a libertarian or whatever.

But he didn't mean that.

And it's like, well, what does anyone mean then?

It just becomes kind of nonsensical.

But in this paradigm, things are very different.

Politicians would start competing on the metrics.

So in order for their number to go up, though,

they have to actually hit the target in reality.

It doesn't matter what they say.

It makes absolutely no difference what they say,

just what the market believes.

So as a kind of crazy example,

a politician could announce that they want everyone to die

and that if elected, they'll kill everyone.

But if traders don't actually believe that,

traders survive, will survive somehow.

They'll kill everyone except the traders.

But if the traders don't actually believe that,

then the death number, that third row from before,

that won't actually move.

Because it's only about,

the asset only pays out much later

when the event is actually measured.

So this is just a different world

where you just don't have to listen to.

I mean, it's possible that,

I mean, in the ideal, my sort of ideal world,

the ideal world of this idea,

the politicians would just kind of put out

some PDF document of their plan

and then they just wouldn't speak again

and then the election would just happen.

And there would just be no speaking at all

and none of these weird rallies or this weird other stuff.

So that's kind of the idea.

But we still haven't addressed these two lingering issues,

which we will now,

which is about how do we get people to,

I'm sorry, this paragraph I didn't read

is about what I was saying about the strategy involving,

how it's different from milk and stuff.

So it's not really that important.

But yeah, you can have,

many of the in-trade markets

had very little liquidity and very few trades.

Some had no trading, literally.

And so we're going to solve that

with this sort of standing army of automatic traders

that I'm going to explain.

And then you have people

try to stigmatize one market and say,

this market is a bad one, this is a bad egg,

don't pay attention to it.

But we can actually make it,

it's the same solution for both points,

we can make it so that,

you can create a market such that X dollars

can always be earned by betting in it.

You know, if the prices are wrong,

you disagree about the price,

you can always extract X dollars out of it,

even if no one else on earth is trading.

And then the plan is to make X like a million dollars or something.

That would be very persuasive to a layperson voter.

Someone living in a swing state,

in like Ohio or whatever,

they're going to be like, really?

There's millions of dollars at stake if the prices are wrong,

but this guy is saying that I should ignore this?

Even though there's millions to be won?

I mean, as much as possible, ideally.

But, okay.

So, yeah.

This is Robin Hanson's response to this critique,

which is that you don't have to get the information for free,

you can buy it,

and you're going to use this sort of sacrificial trade

that loses against all the outcomes at once,

using something called the market scoring rule.

And this presentation is almost over,

there's a million things in it,

but it's winding to its end.

This is the last part.

So, this market scoring rule is very different

from the way trading usually happens.

And I'm just going to go through it.

And I don't know, this is like the first time I've tried to explain it,

but it has many differences.

One is there's no offers, there's no bids, asks.

Stuff just happens immediately.

Let me see, I should put some of my notes.

So, everything just happens.

There's no bids or asks, you just start with someone,

and then a trade happens, and, oh well.

Okay.

Another difference is there's no interaction between buyer and seller,

you just interact with this one fixed thing in a big line.

And there's no limbo.

So, this limbo is this thing here.

So, this is a usual timeline,

it's called a double auction market,

where some people do bids and some people do asks,

and they overlap,

and you have to wait until they overlap for a trade to happen.

So, this is how everything normally works,

probably that you've ever seen in your entire life.

And it's, someone creates this market, and the market exists,

and trader one submits a bid,

and this bid is hanging in limbo.

And trader two sees this buy wall,

and they see the bid,

and they say, okay, I'll take it.

I accept it.

And then, it transacts for both of them.

This person sells, and this person buys.

So, the thing goes here, and the money goes up to there.

But, in this scoring rule thing,

you create a market,

and you put some sacrificial cash up front immediately.

And then, when this person wants to buy,

they can just do so immediately,

without interacting with anyone.

It's just very atomic, it just happens.

And then, the market changes state.

So, the market's gonna change state

in a big line of state changes,

and the latest state will be the real state.

So, it fits really well with the blockchain.

And then, when the person wants to sell,

they just interact with the market directly.

So, they don't have to do anything,

and there's a, this is a weird example,

but to try and give you some example,

some instance of what happens here,

is that the first thing that happens in the timeline

is that you put some money here,

in this sort of seesaw.

So, you have this election that has two states.

Normally, you'd create a market for like a no asset.

In this case, there's only two,

so the no and yes would be exactly the same,

so you'd only have one of them.

But, the point is, you could have a bunch of stuff,

no shares, stacks of paper that are worth a dollar,

if Hillary Clinton is elected,

and worth nothing if she is not elected.

And then, here would be the reverse.

They'd be worth a dollar if she is not elected,

and nothing if she was not not elected,

if she was elected, so.

Anyway, then you'd have these markets here

that have these prices.

Each of them would have their own price,

and you'd swap money for little receipts,

little pieces of paper.

That is how it would normally work.

But here, you start it off,

and you put some money in the center.

The trader, and then the trader could buy some

by putting money on the left side.

And as they buy,

they'll be bidding up the price of no up,

so these are reversed.

It's from zero to 100%,

or 100 cents on the dollar.

So, as you buy, you'll be tilting.

You'll eventually be betting against yourself,

if you put too much money on here.

And you'll be buying something for $1 a share

that could be worth, at most, $1 a share.

But then, if you buy, the price of yes will go down.

This, in Euclidean space, it will go up,

but this price of it will go down.

And traders over here would then maybe buy more shares.

When you put some shares on this at a lower price,

there's like an automatic machine off the side

that's printing out receipts.

It watches people put money and take money off of this,

and it gives them receipts.

If you want to sell, you put receipts back in the machine,

and it lets you pull money off of the seesaw, so to speak.

This is an example

describing the market state.

This is in Microsoft Excel.

If any of you want this, it's on the website,

along with a lot of other examples.

But you start out with zero shares issued of each.

So, like, Hillary wins, Hillary loses.

You start with zero, but it costs 485 units,

45 Bitcoin, $45, whatever.

But it costs that much to create this in the first place.

That's the hippo money in the middle.

And then someone buys.

They have to pay .51, but they get one share.

And it alters the prices.

And then when this person wants to buy six,

they have to pay 3.4.

And so this time is flowing down this way,

and you just keep updating these states.

At the end, no matter what happens,

you can never need more than this.

No matter how many shares are bought and sold,

you'll always have enough.

It's fully collateralized to redeem

when A equals zero actually happens.

42 people have shares somewhere,

and you need 42 units, $42 or whatever,

to pay them back.

So, now the cool thing is,

the math is totally indifferent to the number of states.

So I said before, it's terrible with a lot of states.

But in this case, it's actually completely fine.

And in fact, something grows.

The required amount of capital grows logarithmically

with the number of states.

So you have a seesaw in the example I gave,

but if you wanted to do the cross that I described

with four states,

then you just have some money in the middle,

but it's in the middle of a four-way pyramidal seesaw.

And then this is kind of like a plate on a pin or something.

And you can have as many states as you want,

all with the same initial money in the middle.

And that means that everyone will be able to...

this solves the problems I mentioned before,

because someone will always be able to trade at any time,

and you can have huge amounts of money in the center

that's a small amount of money in the grand scheme of things,

maybe like $1 million or $2 million,

but would be very persuasive to someone.

You say, ignore this market,

but you say, actually, if the prices are wrong,

if the price is off by 5 percentage points or something,

someone could make $100,000, $200,000, $300,000, $1 million,

then people will be like, I'm not going to ignore it,

because there's a lot of money I could earn if it was wrong,

and there's a lot of money that anyone could earn if it were wrong.

So that is...

and as again, this section of the talk could probably go on forever,

but I think it's winding down at this point.

Let me see.

So, yeah, I'll just skip all this.

So, yeah, this is the very last part.

I think two slides.

So it's like, where would you get the money?

But actually, I think when the time comes,

if the money is needed,

like if there really is someone creates the markets

and they're stalling out

because there's not enough trading activity,

I think it would actually be pretty easy to get the money,

because you can draw a comparison

to three other types of donation activity.

So we have donations to presidential candidates,

super PAC donations,

and donations to, like, a charity

that helps keep politicians honest,

for example, the Annenberg Political Fact Check.

And you can see that actually,

just in terms of money raised,

it's unbelievable as much as Bernie Sanders

raised $18.2 million in just six weeks,

Kamala Harris raised $12 million in quarter one of 2019,

Pete Buttigieg $7 million,

Elizabeth Warren $6 million.

The super PACs in fiscal year 2016,

they had $1.79 billion in donations.

That's a ridiculous amount of money.

And just this one tiny little charity

that helps to check what politicians say for accuracy,

even they got $1.16 million in total donations in a year.

You only need to assemble this money

once per important election cycle,

so maybe once every two, four, six years.

You just need to have enough

to be persuasive to most voters,

and most, I mean, I think average household wealth

is, like, I think it's 44K.

I'd say around 50K here.

So if you have a lot of wealth,

you could at least get the attention of many people.

I think you might need

as little as $5 million per four years,

which in the grand scheme of all this,

I think is sort of achievable.

It's also easier to,

this is an easier sell in a couple ways.

I'll skip to number four,

and then go back to three.

But one cool thing is that this is,

you're giving money, and it's 100% efficient.

So normally if you give money

to the Bill and Melinda Gates Foundation,

some of it goes to administrators,

where it's just gone.

You have to pay the secretary or something.

And then you don't know how they spend it.

They say they spend it a certain way,

but you don't really know.

But with this, it would all happen on the blockchain,

not to be cliche about it.

And so all the donated funds contribute directly

to the election reform cause.

There's no overhead.

None of the money goes to me, for example, at all.

And none of it can possibly be wasted.

It goes directly, you have a direct impact.

Another thing is that it's private, uncensorable.

No one can stop you or someone like Roger Ver

or something from donating,

just because they don't like him.

You can add a little message to your donation.

If you want to be unanonymous,

you could say this donation in memory of whatever

and something like that.

So you can do stuff like that.

But another point is this third point,

is that a lot of people are unsatisfied,

but they don't know what to do,

what specifically to do.

And the people who do come up with things to do

often end up, they sort of split the vote in a sense,

which is that you have people who are like,

both parties suck, so I'm voting Libertarian.

But you also have people who are like,

both parties suck, so I'm going to vote Green Party,

because what difference does it make?

Or I'm going to vote for,

I don't remember who Jill Stein is.

Is she the Green Party?

I don't even remember.

But there's another one.

We have a small, at least in Connecticut,

there's a small labor party.

And I think there's a reform party

that's a real party.

Yeah, there's extreme.

So that's the other thing,

is that if you're Libertarian,

there's extreme factions of Libertarianism as well.

There's anarcho-capitalist,

and then there's Ron Paul sort of people.

And then there's many different groups,

and they all disagree.

But with this, it's kind of easy.

You're all agreeing not to agree.

You're just saying,

this money is going to go in the center,

and we don't know which candidate

it will end up helping,

because the process hasn't even started yet.

So you don't have to think about it too much.

You can just think about how angry you are

at the existing options.

And you don't need a third-party candidate to win.

You can just induce the two major candidates

to compete on those metrics.

And you can add as many metrics as you want

because of this dish thing.

You can make millions of rows

and have tons of states,

and it will only be diluting the liquidity log and size.

So it will dilute it somewhat,

but not that much.

And so, yeah, that is the talk.

Yeah, that was a lot of stuff,

but thank you for listening to it.

Thank you.

Yes.

I'd be happy to take any questions.