The AGI economy
Introducing a new blog series
0. Housekeeping: Welcome back!
A new blog name: From March to June I have written a blog series that explores AI analogies. There are still many AI analogies out there and I will continue to occasionally add to that series. At the same time, I want to increasingly focus on more actionable advice for individuals and institutions to prepare for a world filled with AGIs. The new name “Machinocene” is broader and reflects the idea that we are slowly but inexorably moving towards a future shaped by intelligent machines, much like we shape the Earth in the Anthropocene.
A new series: I am excited to be a part of the 2024 Blog-Building Intensive Fellowship Cohort of the Roots of Progress Institute. My main focus in the coming months will be thematically coherent series on specific public policy challenges of AGI futures - starting with the role of human labor in the economy.
1. What the AGI economy series is about
Economists offer a wide range of views on the most likely impact of AI:
Daron Acemoglu predicts a maximum 0.53% increase in total factor productivity over 10 years.
Others have argued that the advent of AGI could lead to extreme economic growth rates like 30 or 40%.
Some economists such as Daniel Susskind are worried about the prospect of near-term technological mass unemployment
Others such as Noah Smith think this is yet another false alarm and we need automation as fast as possible to help workers and deal with ageing societies.
I don’t have a strong view on most of these questions. However, we should assign at least some probability on increasingly autonomous and agentic AGIs having a very significant impact on our economy and we should aim to have robust institutions that can perform well under a range of timelines and scenarios. To stress test our institutions,1 I will focus on one specific structural shift.
Guiding hypothesis: Labor income as a share of national income will decrease in an AGI economy. Correspondingly, capital income as a share of national income will increase in an AGI economy.2
The decrease of labor income as a share of national income is an already occurring trend and it seems plausible that AGI will further aggravate it due to deskilling and cheap substitution of significant amounts of labor.
If the share of capital income rises to more than 50% of national income, we could call this a “post-labor” economy.
2. Why would a shift from labor to capital matter?
Our current political institutions are not set up for such a “post-labor” economy. A few of the key challenges include:
The inequality of non-labor income, particularly from stock ownership, is much higher than the inequality of labor income.
If significant shares of the population will be without substantial labor income due to frictional or long-term, structural technological unemployment due to AGI this would increase the need for public social security spending.
About 50% of government revenues in OECD countries come from labor income through personal income taxes and social security contributions, whereas corporate income tax is about 10%. Meaning, government revenue might not scale in lockstep with the machine economy.
The good news is that it is possible to design economic institutions that perform robustly across both labor and “post-labor” economies. This can reduce the risk of social and political turbulence and helps to work towards a new level of human thriving. At the same time, historically, institutional adaptations have required time, so we should start thinking about them sooner rather than later.
3. Series outline
We will first explore the idea of a “post-labor” economy including some important limitations
Subsequently I will provide overviews of commonly suggested institutional solutions, including a discussion of specific implementation challenges:
Lastly, we will look at the approaches that I consider the most underrated to prepare for an automated economy:
“wind tunneling” in foresight lingo
The capital share of national income can be calculated with a simple formula:
Capital share of national income = annual return on capital * (national wealth per capita / GDP per capita)
For illustration, let’s go through an example with specific numbers:
a GDP per capita of 30’000 $
a total wealth per capita of 180’000$
an annual rate of return to capital of 5%
In this case the capital share of national income is 30%. 5% * (180’000/30’000) = 5% * 6 = 30%. In other words, the 30’000 $ GDP per capita consists of an average of 21’000 $ in labor income and an average 9’000 $ in capital income.