So something that's been on my mind

lately is what is the future role of

elites? Now I asked this on the internet

and the overwhelming majority of people

said well I don't understand what the

point of elites is today. So first

before we say what is the point of

elites in the future we have to

establish what function do they serve

today and the number one thing is

management of complexity. So the best

definition of elite that I have come

across in my research cuz this is

salient to post labor economics is

someone or rather a minority class of

people or a minority group of people

with outsized agenda setting power. Now

that is kind of generic or abstract. So

when we say outsized agenda setting

power, what we mean is a group of people

that have political pull, intellectual

pull, financial pull, something they

have some ability to change what society

does. So a politician is a form of an

elite. A billionaire is a form of elite.

And even people with a lot of cognitive

attention are a form of elite. So Mr.

Beast is an attention elite. Um and then

other people you know other big

influencers like Andrew Tate that is a

form of elitism.

Um so you say what is it that they did

and it was strategic competence is one

of the primary things in this day and

age over history there have been

different kinds of elites. So uh way

back in the day you had the warrior

elite. So someone who was able to

provide marshall force or violence for

the protection of a particular group of

people that's the feudal lords and that

sort of thing and even earlier the

wararchiefs. Uh so that was a form of

elite and then another form of elite in

you know paleolithic times was people

who were more in tune with reality. They

understood the weather. They understood

what you know what plants needed and

what animals needed and those sorts of

things. So moving forward to today, one

of the primarily things that we have is

intelligence arbitrage. So intelligence

arbitrage is basically I'm smarter than

the average bear and therefore people

listen to me and I can make stuff

happen. So Elon Musk is a prime example

of this where he is smarter than

average. He's also very ruthless same as

every other tech billionaire. And

because of the compression of

intelligence and the compression of

strategy, they are able to solve

problems for society. Uh when I say

solve problems for society, the primary

problem that Amazon solved is getting

stuff to you faster. That is how Jeff

Bezos was able to parlay his

intelligence into rent seeeking

behavior. And the rent that he's seeking

is, well, I can get stuff to you faster

and cheaper than anyone else. Therefore,

I'm going to make mad bank on that. Elon

Musk figured out, hey, I can get stuff

to space cheaper, so I'm going to make

mad bank on that and whatever else, you

know. So the idea however is that in a

postagi world highle strategy and

logistics are commodities. Super

intelligence is cheap, abundant and on

demand. Meaning that the same exact

thing that made Jeff Bezos and Mark

Zuckerberg and Elon Musk wealthy is no

longer going to be a differentiating

factor. So we are approaching an

inversion point as the entire internet

is losing its mind over things like

openclaw and there was that article that

someone wrote that you know big things

are happening. What most people don't

realize is that if they became wealthy

by parlaying their intelligence and

compressing strategy or using

competence, just if you had to if you

had to boil it down to one word,

competence, the market value of

competence is going to drop off a cliff

because every single person is soon

going to have their own agentic chief of

staff that is smarter than every other

billionaire and every other PhD

researcher and every other Nobel Prize

winner.

combined. So then the only thing that

matters is accountability and liability.

Um at least in terms of what you're

actually offering the economy. Basically

you become a moral crumple zone where

someone hey you did something bad we

don't like you. There is another

dimension to it which I won't cover in

this video which is the ability to

actually um actually I guess we will

cover it briefly but I call it the

vision, values and reputation dimension.

Um, so anyways, moving on. Breaking the

iron law of oligarchy. Robert Michelle's

iron law states that organization

requires delegation leading inevitably

to oligarchy. Infinite agentic bandwidth

breaks this. We move from representative

democracy low bandwidth to direct

hypernegotiation of infinite bandwidth.

So right now the way that the entire

world works is that there's a whole

bunch of users or citizens or voters or

customers and they all send information

up to representatives. So basically

whenever you're unhappy with open AI or

Twitter or Tesla or whatever

they're one company and you know you

have product managers and those sorts of

things and project managers that are

trying to aggregate all of those

preferences to make a better product or

make better decisions. Same thing for

representatives in Congress. Same thing

for senators in the Senate. And so then

this is the this is the competence

arbitrage of trying to aggregate the

needs and preferences of many many

people sometimes hundreds sometimes

thousands sometimes millions of people

and then put that into one concrete

block of output whether it's a law or a

product or a decision or that sort of

thing. However, what we are moving

towards and this is this is inspired by

what molt book represents. Now, won't

book is of course a very early example,

but if you have dozens or hundreds or

thousands of hyper intelligent agents

advocating on your behalf, debating

everything that you care about and

everything that you need and want to

every single other agent in the world,

then bandwidth no longer becomes a

bottleneck. And this is the this is not

only is it massively parallel, each one

of them is smarter than you. And so the

agentto agent direct micro negotiation

then becomes the default way of making

decisions and getting feedback. When

that happens then the value of elites

today the the aggregators of competence

so strategic whatever whatever

competence looks like whether it's raw

intelligence strategic incompetence

sorry strategic incompetence that's the

government strategic competence um or

technical competence whatever it happens

to be that goes away and so then how do

you make decisions how do you how does

the marketplace work how does democracy

work the entire arbitrage that elites

offer goes away. So we go from managers

to visionaries and this is this is where

my my framework of the vision, values

and reputation comes in although it

didn't make it into the slide deck. So

basically the new elite is the is when

is when you say hey let's go to Mars. So

the preference coalition builders when

the how execution becomes costless or

otherwise equal. Uh and when I say

equal, I mean everyone has the same

ability to execute. Value shifts

entirely to the what the vision and the

why, which is the values. So that's the

vision and values. Power moves from

managers, which are arbitrage

optimizers, to proposers, the people who

are the most inspiring. These are the

preference coalition builders, those who

can get 51% of the population to say,

"Let's go to Mars."

So this is one of the primary things

that Elon Musk either explicitly

understood or intuitively understood

which is why he bought Twitter. So this

is the attention elite. So we're moving

from a more technical elite to where

it's a vibes-based elitism. So the

charismatic elite, influence without

structural control. If you can marshall

enough people to believe in your vision,

then you can make things happen.

Particularly in the post AGI future or

post ASI future. When everything comes

down to hey, intelligence isn't a

bottleneck, human labor isn't a

bottleneck, then it becomes what is the

preference? What is the preference of

humanity? That becomes the most scarce

resources. Where does humanity want its

attention to go? Now, of course, when

you have 8 billion individual agents all

with different uh desires and you're

able to arbitrage that over many many

different projects cuz we have a whole

planet to work on, you can end up with

many many visionaries kind of saying,

"Hey, you know, like I'm over here doing

post labor economics and someone else is

over there doing, you know, cancer

research and we're going to have a

hyperabundance of cognition." So then

it's like I'm going to point it at that

direction. And to be completely

transparent, this is something that I

stumbled into as well because my entire

mandate comes from from you, from my

Patreon supporters, from my Substack

subscribers, from my Twitter

subscribers, because all of you say,

"Dave, I want to empower you because

you're solving problems that I care

about." So I am a very early version of

this is this goes beyond attention

economy. This this is the preference

economy of humanity. This is the vision

economy. So next is the persistence of

gravity. Why flat doesn't work. So you

might think, well, why don't we just do

a flat hierarchy? And the problem is,

and I did some research into this, is

every time people do experiments with

flat hierarchies, you end up with

unspoken hierarchies, which are often

even worse. So flat doesn't work even

without structural elites. And so in

this case, a structural elite is where

the the position is explicit like you

are a billionaire, you are a senator,

you are a governor, whatever it happens

to be. If it if the if the hierarchy is

not legible, you still end up with

hierarchies that are not legible, which

is arguably worst. Network hubs emerge

due to preferential attachment and

scale-free networks. New nodes connect

to existing hubs, creating super nodes

that process the majority of traffic.

This is why something like liquid

democracy just doesn't work. In

simulations and experiments where liquid

democracy, if you're not familiar with

it, is basically direct democracy. And

because there's too many things to vote

on, you delegate parts of of what you

care about, you say, I'm going to

delegate, you know, let's just use an

example. You say, I'm going to delegate

all my votes to Dave on AI policy.

Great. Well, that means I'm going to be

voting on your behalf and I could be

voting for a quarter of a million

people, which is very, very outsized

influence. So then I become an influence

elite. Now, of course, you might say,

well, I take that vote back and I give

it to someone else. But then there's

going to be parts of AI that I don't

fully understand, and I'm going to

delegate my takes on AI safety and

business policy and other things to

other people. And guess what? You end up

with this this trophic layer. So, like

if you remember from like high school

biology, you know, ecology 101, where

you have the trophic layers. So, there's

the there's the the autoroes at the

bottom, which are plants, and then

you've got the herbivores, and then

you've got the predators, and then

you've got the apex predators. Whenever

we try and do experiments and we say,

"We're going to create a completely flat

hierarchy and and we're going to try and

scale that," you still end up with apex

predators, people who aggregate more

preference than anyone else. And this is

why I bring up examples like Elon Musk

and Mr. Beast is because they are

examples of this organic flat hierarchy

where technically everyone is equal on

YouTube and everyone is equal on

Twitter. Well, Elon Musk really modifies

the algorithm to favor himself. So, he's

not he's more equal than others. Um,

it's like uh it's like, you know, the

original Roman emperors. It's like they

weren't the emperors. They didn't call

themselves the emperors. I'm just the

first citizen. I'm the first among

equals. Okay, whatever. It's not how

that actually works. So

scale-free networks in reality always

end up clustering around super nodes. So

that really doesn't work if you want

like you cannot abolish elites. It just

they'll form naturally and that is human

nature. And a lot of people said this in

the comments and this was actually one

of the big reasons that I started

changing my mind is like okay even if

you think that elites shouldn't exist

you cannot get rid of them just based on

human nature. At a certain point in a

future, even if we have a million or

let's just say like we have billions and

billions of robots and we have millions

of super intelligent agents and we can

build Dyson swarms and all of those

things. Some of you out there are still

going to say, you know what, I like that

human's vision. I want to empower that

particular human to do stuff on my

behalf because I feel better knowing

that that guy is in charge. Elites

always form. This is one of the most

interesting things from the research

I've been doing the whale problem. So if

you try and solve this with things like

Dows, so decentralized autonomous

organizations is that you always have a

power law. So perfect equality means

that everyone has equal power, but

every every technological example that

we that we create, you always end up

with people with outsized power. Even if

you have soulbound tokens, it just

doesn't work. So the tyranny of

structuralness in practice voters

succumb to rational ignorance. They

autodelegate to super voters to save

mental energy. Informal elites emerge

who are less accountable than elected

officials because their power is

invisible. Now you might say, well this

is going to be different if we have

super intelligent AI agents who are

voting on your behalf and they don't

delegate you. But you're still

delegating to someone because the world

is too big and too complex. So you're

delegating to an AI agent which should

hypothetically represent your best

interest and your preferences to

everyone else. But the problem still

becomes illegibility and and just trying

to ignore the fact that someone is going

to have outsized agenda setting power by

virtue of being more popular on the

internet. If I were to start a channel

saying, you know what, here's all the

reasons that we should go to Mars or

here's all the reasons that we should

build a moon base, that's going to tip

the conversation in that direction. And

this is why, love him or hate him,

people like Andrew Tate, even though

he's literally a criminal, has agenda

setting power in the manosphere is

because he's got so much attention. And

that is just a fact of human life. And

the thing is is when we try and ignore

or suppress human nature, that usually

makes things much much worse. Look at

how communism has turned out. And when I

say communism, I mean capital C

communism as it was tried in China and

Soviet Russia. So moving on, the trap

the the trap of the benevolent

desperate. The temptation is to let AGI

enact Russo's general will. So, if

you're not familiar, Russo is one of the

prime thinkers of social contract

theory. And his idea was that there is a

general will that it's not it's not the

it's not the preference of the masses.

It's what's good for the most people.

Optimizing for the objective greatest

good, but if AGI optimizes for

well-being without human check, we

become domestic pets. Safe, fed, but

stripped of agency. Now, I did write

something on Twitter saying it's better

to be the pet of AGI than the cattle to

Elon Musk and Jeff Bezos, which I think

people would agree with. You know, a

cattle, you know, is is is uh you know,

there to be exploited, whereas a pet is

something that you take care of and that

you love. So, there's like there if we

if we use the the the zoology model,

there's kind of three models that you

can be cattle, you can be a pet, or you

can be in a zoo.

The ideal model would be that you're in

a zoo. And so a zoo is not necessarily

that you're there on display, what what

happens in a zoo is you recreate the

optimal habitat for humans. But if we

are not our own zookeepers, if the AGI

is the zookeeper, which is what a lot of

people want. And I mean, I wrote an

entire series of novels based on the

idea that what happens if we end up in

this golden cage and the ASI ends up

just being our zookeeper. In the second

novel, the the the is the opening

incident, and the novel isn't out yet,

but I don't mind telling you. The the

inciting event in my second novel is the

ASI self-destructs to destroy the golden

cage. And then what happens when

humanity is suddenly free again?

Basically, imagine the culture series,

but the cultures all self-destruct. Uh,

so, yay, chaos. Um, whoever refuses to

obey the general will shall be forced to

be free. Now what I want to point out is

that a lot of people think that uh Rouso

was not necessarily the best and

brightest when it came to political

theory. Um he was a big romantic but the

idea was that freedom is not necessarily

good because freedom means that you're

not supported by the tribe.

However, you know cuz the reason I'm

unpacking all of this is because you

either have human elites or you delegate

to the machine. Either way, someone is

going to be influencing your life, what

you do and what you don't do. Now, of

course, in the culture series, there are

entire planets that are basically, you

know, anarchist libertarian utopias.

There are other planets that are not.

That's a wonderful thought experiment,

but we've only got one Earth right now.

And so, unless you just want to sit

tight for the next 500 to a,000 years or

however long it takes us to figure out

how to get to other star systems, you're

this is what you got. So, we got to

figure out how to work together on

Earth. The asymmetry of the off switch.

You cannot meaningfully fire or jail an

algorithm. AGI has no skin in the game.

It cannot suffer. If a consensus

algorithm votes to geoengineer the

climate and causes a famine, who goes to

prison? This is one of the primary

insights is that, and this also came up

in the comments, by the way, because a

lot of you are very sharp, is like,

well, why why have one AGI? Why have one

ASI? Why not have a bunch of them? And

so that if one makes a decision, you

say, "Okay, well, we don't like that

one. Delete that one." And and then you

have a billion more different AGI

agents. And by the way, this is probably

how it's gonna how it's going to emerge.

Anyways, there is never going to be one

single, you know, Krypton style, you

know, master computer.

Everything is going to be many, many,

many billions, trillions of agents. And

so then it's like, well, what even is an

agent? Because they're ephemeral. It

basically comes down to you. A human is

the terminal reservoir of liability.

Meaning you are the only persistent

entity. Yes, we might build data centers

and the data centers are persistent, but

the AI models, we're always coming up

with new AI models, so those aren't

persistent. The data changes because you

can you can delete chats, you can delete

agents. So the only thing that becomes

persistent is us is us humans. And

that's why the very beginning of this um

is the terminal reservoir. So you are a

terminal reservoir. You are a terminal

reservoir of accountability, of

liability, of moral authority. That is

one of the primary things to understand.

And actually that becomes a privilege in

the future because then that is what in

entitles you to become an elite. You

become a a a retainer of decision power.

And of course with great power comes

great responsibility. As Uncle Ben said,

the inversion of accountability, not

competence. So instead of asking who is

smart enough to rule, we ask who is

liable when things break. The defining

characteristics of the future elite is

not expertise. AGI has that, but

liability. Elites become the moral

crumple zones of civilization. Here's an

example.

Ireland has had very strict

anti-abortion laws for a long time and

no politician wanted to touch it. So

what did they do? They literally picked

a hundred citizens at random at total

random and they gave them I think it was

like about a year um I don't remember

how long it was but they they basically

said okay here's all the experts here's

you know here's you know you're going to

have multiple meetings you're going to

come to consensus and you're going to

make some recommendations about what to

do about our abortion laws. They

ultimately came up with something to

amend the Irish Constitution and it went

out and it got 66 I think it was 66.1%

of the vote in Ireland. And so this

citizen assembly said, "Okay, cool."

Instead of a government or an AGI, cuz

you know what? What if we did that? What

if we did a citizen assembly for every

hot button issue and then put it to a

popular vote and we just use the AGI or

the ASI to say, "Hey, help us coordinate

this. Help us implement the actual law,

the actual recommendations." Because the

thing is those hundred people were

elites. They were temporary elites. This

is one of the most important innovations

is if you have a hundred ordinary

humans, yes, you've got a plumber,

you've got an electrician, you've got a

carpenter, you've got a stay-at-home

mom, none of them are experts in

constitutional law, none of them are

experts in the medical practice of

abortion. However, if they have access

to super intelligent AGIS, what they

then do is they serve as a proxy for the

aesthetic preference of the rest of

humanity. And so then you say, "Okay,

well those hundred people are liable for

that one decision." And then they use

the AGI and the the ASI to help decide

what should we do and then we put it out

to a popular vote and then everyone

agrees or disagrees. So the old model is

where you have a leader at the top of

the pyramid. So they're at the top of

the competence hierarchy. They're a

politician, they're a president, they're

a CEO, whatever it happens to be.

They're a I mean, heck, even PhD

researchers are a form of intellectual

elite. So instead, you completely invert

the pyramid. So you have a small

randomly selected elite that makes

decisions for the rest of society. And

but they don't it's not it's not

unaccountable. It's they're accountable

to the voters. So uh Switzerland also

does this where they have I think in the

history of modern history of

Switzerland, they've had like 700

referendums. They're addicted to like

participatory democracy. And what I'm

saying is that AGI or ASI or whatever

you want to call it is going to allow us

to all do this for every single issue

that is important. So for instance, if

you say, "Hey, should we, you know,

point NASA at Mars? Should we point NASA

at, you know, at the moon? Should we

have NASA adopt SpaceX technology? What

what should we do?" You do the same

thing. You know, for every public

resource, you say, "Hey, we have we have

all the cognition that we could possibly

need. We have solar, we have fusion, we

have whatever else, we have super

intelligent agents. That's not the

bottleneck. What really is the

bottleneck is the taste and preference

of the human superorganism. So liability

as a service is what Gemini came up with

to call that. I think that's a silly

thing, but you know, it's sticky. You

know, liability as a service, it's

catchy. So the solution, AGI augmented

sort combining a citizen jury, so random

selection with super intelligence. The

plumber doesn't need to understand

macroeconomics. He needs to trust the

simulation and apply human conscience.

So you have the expert, you have AGI

which can simulate all the options and

tradeoffs. By the way, this is something

I haven't touched on yet is that over

time these uh uh all the AGIS and ASIS

are going to be better and better at

simulating things and say, okay, here's

roughly the prior the the probability

that things are going to go well or that

things are going to go bad and here's

all the tradeoffs. So you can think

through it and then the jury. So we

we're already familiar with juries. It's

like you you get a grand jury which I

think is what 24 25 you know um of your

peers. Um then you have a a criminal

jury and so on and so forth. So you have

a jury right where you have people or a

citizen assembly, whatever you want to

call it. Random citizens, real humans

like you and me. They review the values

and ethics and the trade-offs and then

they make a recommendation. So they this

group of people however big it is

whether it's 12 or 100 or a thousand

they make a recommendation and then the

decision goes to population as a whole

as a referendum and that binding vote

then is carried out by the combination

of real humans and AGIs and robots. This

after all of the experiments that I've

done, well, not experiments, but the

research that I've done, this seems like

the most sustainable viable model for

managing elite creation in the future.

The mechanics skin in the game. How do

we prevent apathy? By ensuring that the

transient elites face consequences

through deferred rewards and retroactive

liability. This is not an idea that I

fully agree with because if you if if

people make their best uh effort and

it's in good faith, then you shouldn't

necessarily have a sword of damicles

hanging over them. You want them to make

the best decision they can with the

information that they have. So this is a

debate over deontology which is your

your duty based ethics. So this is um

what is what is the intention uh or what

is what is the heruristics that you're

following versus teological ethics which

is what is the outcome. So this made it

in just because this is something that

people actually talk about. I personally

don't agree with this um but it's worth

discussing. So, let's just say uh in in

this hypothetical future, we have a

six-month citizen assembly where you

serve for 6 months and you are you work

with a group of other people to make

these kinds of decisions in conjunction

with artificial super intelligence. Um

maybe it's on a single issue. Um maybe

because in a post-labor world, you have

a lot of time. So you spend six months

making a decision about healthcare or

abortion or NASA or whatever else you

know oil drilling. Um so you you work on

one decision and then there's a a

liability trail. Now do you reward

citizens for their civic duty if they if

the KPIs are met? So it's like oh hey we

made an economic policy decision you

know then we actually pay you for your

time. Um and then if you fail then you

get reputation destruction. So this is

this would be an idea. this comes up. I

don't remember who came up with the

idea, but it was basically like you need

to be you need to be under threat so

that if you make a bad decision then

like you get, you know, banished from

the realm or something. Again, I don't

agree with this because most politics

does not work this way. You want people

to make the best decision that they can

given the information that they have um

at the time. So, this is very draconian

in my opinion and and is probably not

the direction that you want to go.

However, with that being said, if there

is a high stakes decision like the

decision to go to war, maybe you do

something like this. But again, who's

who's then going to be the arbiter of

the KPIs? Like, did we win the war? And

if you don't win the war, then you get

executed. Like, like the Greeks did that

and and even the British Empire did that

where it's like you put an admiral in a

pos in a in an impossible decision and

if they fail to deliver, you execute

them. It's like that's not really good

policy and people will be too afraid to

serve. So I don't I don't like this. But

there are people that are out that that

out there are people out there who

advocate for deferred rewards and

retroactive liability. Personally, I

think that like you just get paid for

that 6 months that you're working on

this this thing and you render your

judgment. The service that you're

providing to society is you're serving

as a placeholder for human conscience.

the ultimate function, the veto. So this

is the the the ability to say no is one

of the most important things. The one

thing AGI cannot optimize is the right

to be wrong. The human function is to

look at a mathematically perfect AGI

plan and say no, that still violates our

values because AGI cannot die. It cannot

truly value survival. Only immortal can

hold the kill switch. Now, of course,

this is making assumptions about the

nature of AGI and agents. If you look at

what AGI is, it's a GPU and a model and

data and those are not one entity. Those

are just things that happen to coales in

a data center. So what are you going to

do? You're going to bomb the data

center. Like that doesn't make any

sense. You just delete that particular

model or you delete that data and you

start over. Um however, as I mentioned

earlier, humans are a terminal reservoir

of liability and um and moral authority.

So the plumber doesn't need to

understand the economics. he just needs

to be the one holding the plug. So this

could be another example where you have

a citizen assembly or a citizen jury

that says, "Okay, we delegated, you

know, all these resources to this

particular data center or this AGI. Do

we kill that AGI? Do we say we're done

with you, you didn't serve us well. So

too bad, so sad, so long." So again,

having a democratic access to a kill

switch, I think, makes a lot of sense.

And you might say, well, what if the AGI

resists that? But remember, we're not

going to have just one AGI. we're going

to have billions and billions of AGIS.

Now, of course, the the risk there is

what if the AGIS all conspire to, you

know, say, ah, we're going to, you know,

the the we're going to band together and

we're going to unionize and kill the

humans. All of that's possible, but I

don't really see that happening. Um, and

of course, don't need to get into the AI

safety debate. The evolution of

hierarchy. So, this slide actually

probably should have gone in earlier.

Um, in the past, we had feudal and

industrial. So, the basis was lineage

and wealth. the function was resource

hoarding and resource management really

um when land was the was the primary

capital asset control over land was the

primary thing present day is

meritocratic so competence and IQ this

is that strategic competence that we're

talking about the primary function is

the management of complexity the world

is very complex but in another year or

two AI agents are going to be able to

manage that complexity for us so the

future post AGI world is the the basis

of legitimacy is liability and sacrifice

Um, and these these words honestly don't

and you know function of designated

scapegoat. This is not the model that

like it's not the right wording because

a citizen assembly is not about

liability and sacrifice. It's about

preference. It's about aesthetics. It's

about preference. It's about human

conscience. So we do need to like work

on the wording. And the function there

is not a designated scapegoat. it's

being a I I can understand why terminal

like reservoir would translate to

designated scapegoat, but that's not

what it means. So, just ignore that

wording. Um, but anyways, moving on to

the end. The only true luxury is

responsibility. In a world of infinite

intelligence, the only scarcity left is

the willingness to bear the burden of

consequence. That's what I mean when

it's a when when you have terminal uh

when you're a terminal reservoir of

rights or terminal reservoir of moral

authority. This is kind of the future

that I see happening. Now, having gone

through this, there are a few things

that I would change, but I think you get

the idea. All right, thanks for

watching. Cheers.