Seth Walsh
Iconoclast
Contributor
- Joined
- Jan 12, 2020
- Posts
- 11,161
- Reputation
- 22,826
AGI has been reached.
1/ If AI destroys the free market for human labour, the entire distribution system changes.
Today, labour receives income because firms need scarce human capability.
If that scarcity disappears, wages stop being an economically necessary share of production.
Returns flow to capital.
2/ This is the key distinction people keep missing.
AI replacing “jobs” is not mainly about unemployment.
It is about removing labour as a scarce factor of production.
Once human cognitive labour can be reproduced at near-zero marginal cost, its market-clearing price tends toward zero.
3/ Capital does not have the same problem.
The productive system still requires ownership of:
• compute
• energy
• models
• data
• physical infrastructure
• intellectual property
• distribution
• land
• financial capital
Those remain scarce.
So they retain claims on output.
4/ In the old regime:
Capital + scarce labour → output.
Both factors bargain over the surplus.
In the new regime:
Capital + AI → output.
Human labour becomes optional.
The bargaining structure disappears.
5/ That means “employment” can stop functioning as a market institution.
Companies will no longer need millions of workers simply because production requires them.
A small number of humans may remain around the machines.
But who gets those positions?
That becomes a very different question.
6/ When a resource is economically unnecessary but socially valuable, allocation becomes political.
Employment becomes less:
“Who can produce the most value at the market wage?”
And more:
“Who is permitted access to the remaining high-status institutional positions?”
7/ Those positions could become extraordinarily scarce.
One trusted human directing 200 AI agents may replace an entire hierarchy.
So the surviving human roles become control rights:
Who gets the credentials?
Who gets the model access?
Who gets the budget?
Who gets signing authority?
Who gets inside the institution?
8/ At that point, social class matters far more.
Not necessarily through some explicit conspiracy.
Through ordinary institutional selection:
elite schools
family networks
reputation
social trust
cultural fit
geography
inheritance
connections
political access
When there are 100 good jobs instead of 100,000, allocation gets much less anonymous.
9/ A competitive labour market can partially discipline nepotism.
If you desperately need 10,000 engineers, you eventually have to hire people who can engineer.
If you need twelve humans to supervise an AI stack, you can afford to choose people who are already inside the trusted social network.
10/ This creates a strange world.
Material production may become vastly more efficient.
But access to productive income becomes vastly more concentrated.
GDP can explode while the economic relevance of ordinary labour collapses.
There is no contradiction there.
11/ The political fight then moves upstream.
Not:
“How high should wages be?”
But:
“Who owns the productive capital?”
“Who receives transfers?”
“Who controls the AI systems?”
“Who gets access to institutional positions?”
“Who has claims on the surplus?”
12/ In that regime, inheritance becomes more important than employability.
Equity ownership becomes more important than credentials.
Political power becomes more important than labour bargaining power.
And social class becomes more important precisely because the market mechanism that once rewarded scarce labour has weakened.
13/ The deepest mistake is assuming that technological abundance automatically produces broadly distributed abundance.
Production and distribution are separate problems.
AI may solve the first.
It does not solve the second.
14/ If labour is no longer required for production, there is no economic law saying labour must continue receiving a large share of output.
That share will have to come through ownership or political allocation.
15/ The end state is therefore not necessarily “everyone has an AI job.”
It may be:
Capital owns the machines.
A tiny trusted layer operates them.
Everyone else receives some politically determined claim on the output.
And the remaining thing called “employment” becomes increasingly a mechanism of status, access and class allocation rather than a true market for scarce human labour.
@Macan @mirinmirinmirin @chang cypionate @Jimcel
1/ If AI destroys the free market for human labour, the entire distribution system changes.
Today, labour receives income because firms need scarce human capability.
If that scarcity disappears, wages stop being an economically necessary share of production.
Returns flow to capital.
2/ This is the key distinction people keep missing.
AI replacing “jobs” is not mainly about unemployment.
It is about removing labour as a scarce factor of production.
Once human cognitive labour can be reproduced at near-zero marginal cost, its market-clearing price tends toward zero.
3/ Capital does not have the same problem.
The productive system still requires ownership of:
• compute
• energy
• models
• data
• physical infrastructure
• intellectual property
• distribution
• land
• financial capital
Those remain scarce.
So they retain claims on output.
4/ In the old regime:
Capital + scarce labour → output.
Both factors bargain over the surplus.
In the new regime:
Capital + AI → output.
Human labour becomes optional.
The bargaining structure disappears.
5/ That means “employment” can stop functioning as a market institution.
Companies will no longer need millions of workers simply because production requires them.
A small number of humans may remain around the machines.
But who gets those positions?
That becomes a very different question.
6/ When a resource is economically unnecessary but socially valuable, allocation becomes political.
Employment becomes less:
“Who can produce the most value at the market wage?”
And more:
“Who is permitted access to the remaining high-status institutional positions?”
7/ Those positions could become extraordinarily scarce.
One trusted human directing 200 AI agents may replace an entire hierarchy.
So the surviving human roles become control rights:
Who gets the credentials?
Who gets the model access?
Who gets the budget?
Who gets signing authority?
Who gets inside the institution?
8/ At that point, social class matters far more.
Not necessarily through some explicit conspiracy.
Through ordinary institutional selection:
elite schools
family networks
reputation
social trust
cultural fit
geography
inheritance
connections
political access
When there are 100 good jobs instead of 100,000, allocation gets much less anonymous.
9/ A competitive labour market can partially discipline nepotism.
If you desperately need 10,000 engineers, you eventually have to hire people who can engineer.
If you need twelve humans to supervise an AI stack, you can afford to choose people who are already inside the trusted social network.
10/ This creates a strange world.
Material production may become vastly more efficient.
But access to productive income becomes vastly more concentrated.
GDP can explode while the economic relevance of ordinary labour collapses.
There is no contradiction there.
11/ The political fight then moves upstream.
Not:
“How high should wages be?”
But:
“Who owns the productive capital?”
“Who receives transfers?”
“Who controls the AI systems?”
“Who gets access to institutional positions?”
“Who has claims on the surplus?”
12/ In that regime, inheritance becomes more important than employability.
Equity ownership becomes more important than credentials.
Political power becomes more important than labour bargaining power.
And social class becomes more important precisely because the market mechanism that once rewarded scarce labour has weakened.
13/ The deepest mistake is assuming that technological abundance automatically produces broadly distributed abundance.
Production and distribution are separate problems.
AI may solve the first.
It does not solve the second.
14/ If labour is no longer required for production, there is no economic law saying labour must continue receiving a large share of output.
That share will have to come through ownership or political allocation.
15/ The end state is therefore not necessarily “everyone has an AI job.”
It may be:
Capital owns the machines.
A tiny trusted layer operates them.
Everyone else receives some politically determined claim on the output.
And the remaining thing called “employment” becomes increasingly a mechanism of status, access and class allocation rather than a true market for scarce human labour.
@Macan @mirinmirinmirin @chang cypionate @Jimcel
