I'm going to talk about three different things, they're all related, and I'm really just going

to mention the first two and then get into this third one.

So the first and most important thing is sort of some like emergency discourse kind of repair

type concept.

So an objective function is something that you want to achieve.

So if you don't agree on that first, the conversation is a waste of time after that.

So I think there are a lot of people here, you know, this is a very diverse community.

Some people are interested in cheap transactions.

Other people want to like single-handedly overthrow most of the world's governments.

Some people want to like timestamp documents.

So if you can't agree on what the objective is, everything you could say after that might

be true and you could still get a different conclusion and everything would be a disaster.

So I just wanted to mention this in two slides, this is the second one.

So related to the block size, if you, if someone hasn't stated what they think the

block size does for Bitcoin, then you can't even tell if what they said is related to

improving Bitcoin according to them, let alone whether you even care if what they said was

right because you even agree with the purpose that they expressed.

So secondly, if you're going to state conditions under which the block size should increase,

you should also be able to say state conditions under which it should decrease.

They're actually the same question because either it's related fundamentally to some

principle or it's just kind of like what you think today and tomorrow it could be something

else or if you went to a different person, they could think something else.

So that's, I just really wanted to mention that because I think a lot of misery, and

I hope it's not too patronizing because, you know, professional philosophers struggle with

this kind of reduction to first principles a lot and I think a lot of misery has basically

been the result of this and nothing else.

So I also wanted to briefly mention something that I wrote a blog post about very recently.

So not agreeing on the objective function is almost as bad as not agreeing on the constraint.

So decentralization is something that I think we would all agree is sort of this constraint

and we need to keep it at some acceptable level.

So I don't really actually want to go into this because there's not a lot of time, but

basically my conclusion is that if you, you need to run a full node in order to know whether

or not you've been paid or you're asking someone else if you've been paid and so you have to

trust them.

So this process of using money becomes more decentralized when it's easier and cheaper

to run a full node and I think, so this is like the scalability comment I wanted to make

is that I think that since it's an engineering requirement that Bitcoin be kind of above

the law.

So it's not like you don't have to want to break the law or anything, but if you acknowledge

the law as an authority, it is no longer peer-to-peer.

So everything that's peer-to-peer, BitTorrent, Bitcoin, whatever, it all is above the law

just by design.

So it's not like, it doesn't have to mean anything politically, but my point is that

I think the real constraint is actually not upstream bandwidth, but I think it's actually

Tor bandwidth so that things remain private and I think it would be very cool if someone

could figure out how to do metering of the bandwidth using Bitcoin itself and with that

being the case, you would have actual price for this bandwidth and that would be the last

abstract thing and you could add them all up and then you'd have an actual number for

measuring decentralization.

But since this is part of the communication block and I really have something important

to say about this concept of peer-to-peer governance.

So I think we can all agree that governance is more important, or at least it affects

the software, the causality kind of like goes up through this pyramid.

The governance is the only thing that can change the protocol's rules and so it's actually

even more important than the operation of the protocol or the software that uses the

protocol.

It's sort of, it's more important.

It's really, really difficult to figure out how to do this and how to communicate this

information given that everyone is going to be an equal peer and I really think that

no one has ever done this before because even something else that's peer-to-peer like BitTorrent,

it doesn't really matter if you fork BitTorrent and you create a new BitTorrent client.

It doesn't affect the people who are already using their old BitTorrent client at all.

So it's like they're all soft forks when it comes to BitTorrent or basically anything

else but with money it's a big deal because what other people do affects you and so we're

kind of like all in this together whether we like it or not.

So you need some kind of governance thing, yet with this peer-to-peer thing it's going

to be very difficult.

So I'm suggesting that we can use markets to inform this governance process.

So it's not like a government, like you don't have to do this, but it's like what should

the government do, governance.

So markets can help us inform this governance process a lot, I think.

So first of all, they're totally peer-to-peer and like permissionless and decentralized

just in their nature because all the information happens locally, but what I really wanted

to explain is that markets transform subjective values into objective information.

So that's a very helpful thing.

When you, in capital markets like the stock market, you make a trade, you have contributed

knowledge to the market.

So you think the current price is X and you change it to Y and you think that the X was

wrong and you think that Y is right.

But more importantly than this knowledge that you contribute, you know that anyone could

have changed the price to anything, but they did not.

It's you who's changing it from X to Y.

So you think that you're actually the number one expert in the entire world because anyone

could have done this thing and picked up expected value and made some money in expectation,

but they did not and you are doing it instead.

So it's like you're really saying that you're the expert.

So this is sort of like a maximization of the signal-to-noise ratio in a way and market

prices are constantly and unanimously acceptable.

So at all times, to all people, the price is an acceptable measure of future reality

because if this is ever not the case, someone can trade and introduce something that, introduce

an edit to the forecast that nets them money.

But most important of all, the real purpose of all this is that prices are what game theorists

call common knowledge.

So if I know the price of something is K and you know that it's K, I mean we can just look

at whatever price we're talking about and we'll both know that it's K and everyone will

know that everyone knows that it's K and everyone will know that everyone knows that

it's K ad infinitum, et cetera.

So common knowledge enables something called free coordination.

So since everyone is on the same page, it's very easy for a group of people to just switch

from one thing to another thing even without any kind of leader being involved.

So that's a really important thing.

So I kind of think of it as actually it's not sort of like you have these proofs in

cryptography that you know like a certain number or something.

But I see prices as kind of like an expertise proof.

You can just look at them and everyone knows that this was an accumulation of a serious

quantity of expertise.

And so that can be very helpful to us.

And it's not commonly known that markets can be about anything.

They don't need to be about capital.

They can be about these events which are today commonly called prediction markets.

So you have these like partitions of mutually exclusive states.

And so on the left here, one of these things is paying a dollar.

Only one of them because Hillary Clinton is either going to be elected or not elected.

And on the right, it's basically the same question.

It's electoral college votes in the United States.

There's 538 electoral college votes for president.

And you can just instead of being, you know, it's like zero, one, you can just split this

thing so you can get a continuous measure instead of a discrete one.

So this has some costs even though it's only like an information thing.

So a cost that it does not have is you're not like sacrificing any of your autonomy

or anything.

You can still decide to ignore this later.

But it does have some costs.

And so the cost of it, you need an oracle, you need this market infrastructure, you need

traders.

And so as was mentioned, I have my own project Truthcoin aims to eliminate the oracle completely

and make it peer-to-peer.

But it's highly experimental and requires peg sidechains which don't currently exist.

So you could make a very, very, very simplified version instead which is sort of what I'm

proposing with this federated idea.

And you could have these multi-sig functionaries.

So this is kind of like a tradeoff, right?

So I'm saying you're picking like something like seven people to just type in five or

six numbers later.

You need to type in, you know, what the objective function ended up being.

But they don't need to use any expertise or anything like that.

They just type in something that we all know and you're just trusting them to do it honestly.

So this is a tradeoff like anything else.

So like for example, the SHA-3 competition that we just heard about, you had to have

experts kind of administer.

That was like endorsed by like the federal government and there was like a hierarchy

of expertise and the experts were sort of in charge.

But with this, you know, you don't have experts.

You have these representatives who you're only picking them based on how likely they

are to do this job.

So you're not saying they're smart or that you like them or anything.

You just think that they're likely to type in these numbers honestly, possibly because

they own Bitcoin businesses that would lose future business or something like that.

So the problem with the experts thing is that it's a circular definition and this is a very

diverse community and it's like who is going to resolve a dispute if you think someone

is an expert and someone is not.

But it's much easier, it's much more objective to pick the representatives.

The other thing is that there's this tradeoff between meta-deception and deception.

So the expert, you might pick these seven people or these three in this case and they

might deceive you, but you would know that you had been deceived because they had these

numbers and they just typed in numbers that were different.

So you would know that they deceived you, whereas today I think we're just kind of confused

in general and it's sort of like we don't know what we're confused about or who is actually

more confident or more of an expert about a certain thing than someone else.

And so it's just a tradeoff like anything else, but this is the bad news.

So I don't really have a lot of time to go into this, but I think people, the other two

things, so the market infrastructure is actually very, very, very easy.

You don't need to do this thing with having everyone submit bids and asks into like a

big order book and like waiting for them to overlap or something like that.

You can actually have something called a market scoring rule, which just has a single state

and then a single person just atomically updates the state and it's only them.

So it's just like passing a single Bitcoin around, only faster.

But the way this works is that there's this formula that you have these mutually exclusive

states and you have the formula that relates the total number of outstanding shares to

an account, the money that the market owns, and people propose updates to the share state.

If you want to buy shares, you propose an update that has more shares.

If you want to sell, you propose an update that has lower shares.

You do this on your own computer.

You use this tiny formula.

You can actually change the formula around a lot, but you use some formula to relate

these things.

And if you pay the difference, then it's valid.

If you don't, it's not valid.

So you can buy.

It issues the shares.

You can do buying.

You can do selling.

On the right, I have this first difference.

So time is on this vertical axis and on the right, I have this first difference.

So this red guy at time equals nine bought 18 shares for 12.86 something, and then this

green guy at time equals 12 sold 21 shares of that right, more right state for 2.7.

But my point is just that this does everything you would want and it's literally one formula

and it just doesn't get any simpler than this.

So that would not be a lot of work and I already have code that does that anyway.

So the other thing is getting the users involved and you might think that you would just create

two different markets and have an objective function for one, an objective function for

another and you would then just look at them and see which price is higher because this

is a representation of the future reality if you do a certain thing like that.

But you can actually go way crazier than that by combining them and now you've got four

mutually exclusive states.

You get a lot of stuff for free.

You get things like decision insurance so you can just set up something that pays you

a ton of Bitcoin if something that you don't like happens.

So if you're interested in, you want to see the block size increase, you could bet that

it would not increase and then after you're disappointed at least you get a lot of money.

So it's sort of like hedging yourself, something like that.

You can do that, you get that for free as just a result of this, you know, summing these

things.

You also get, depends on, you can bet on the objective function itself vertically.

But the cool thing is that for every n dimensions you have here, you get n minus 1 relationships

and that lets things get really fancy because I don't really want to go into it because

it takes a lot of time but you can basically invest a total of one Bitcoin and get a total

of one Bitcoin back if things didn't go your way.

So if you really wanted the block size to increase, you could do this thing where if

it didn't increase, you'd get a Bitcoin back.

But if it did increase, you would get something equal, the entire portfolio that you invested,

the one Bitcoin would be worth the Bitcoin return between now and the time that you made

the investment.

And you can go even crazier than that by multiplying again by the exchange rate and you can put

that entire process in cash.

So believe it or not, you can actually start with $500 US and end up with at least $500

or more, even if the Bitcoin price collapses, as long as it doesn't go to zero, it's actually

the log price, so that's why that doesn't make sense.

You can start with $500 cash and you can get it all back if someone screws up the decision

according to you and if the decision goes the way you thought it should, you'd get something

that covaries with the objective function.

So you basically always have, tons of people have an incentive to participate in this market

and you can even charge these people a little bit of money for this insurance service and

this money can pool up as liquidity, something called liquidity sensitive market scoring

rule and then there's like a big pot of money for ideally like researchers to kind of like

start fighting over and then as they fight over it more, the market becomes even more

liquid and that's how it might work, it might work better.

So I'm just going to really quickly go over the benefits, the Oracle is a slightly annoying

cost, but the main benefit is that you can use the price as an objective function and

the price kind of catches everything, right?

It's like everything you would want, if Bitcoin is useful or whether or not it's different

from something else that we already have and so the price is kind of like this big funnel

that catches everything.

But even if you wanted to have lots and lots of different objective functions, another

benefit is that you can just keep redoing this, it's very simple, you just need the

Oracle to type in these things later.

Paul?

Yes?

I think we're hitting overtime.

Oh yeah.

For this 15 minute presentation.

That's fine, yes.

Thanks for cutting me off.

It was awesome though.

Loved it.

Okay, thanks a lot.

And people remember to, this point is really important about agreeing on an objective function.

Paul?

It's time to go.