🧵We’re asking the wrong question.
The question is not: can AGI plan the economy?
The question is: is our use of AI *already* corrupting the price signals we depend on to coordinate civilization—including the ones guiding how we *build* and *scale* AI *itself*?
#AI #economics
The question is not: can AGI plan the economy?
The question is: is our use of AI *already* corrupting the price signals we depend on to coordinate civilization—including the ones guiding how we *build* and *scale* AI *itself*?
#AI #economics
https://twitter.com/dwarkesh_sp/status/1912937216852336658
1/ The seductive idea:
AGI + Big Data + Supercompute = Perfectly Planned Economy
This isn’t just wrong—it’s economically suicidal.
Oscar Wilde warned of knowing "the price of everything and the value of nothing."
AI threatens to automate that blindness—at planetary scale.
AGI + Big Data + Supercompute = Perfectly Planned Economy
This isn’t just wrong—it’s economically suicidal.
Oscar Wilde warned of knowing "the price of everything and the value of nothing."
AI threatens to automate that blindness—at planetary scale.
2/
AI central planning premise (oversimplified):
1) Information = anything that can be digitized and stored on hard drives
2) Big Data = AI’s fossil fuel
3) Supercomputers + Big Data → AGI
4) AGI = “solves” the economy
AI central planning premise (oversimplified):
1) Information = anything that can be digitized and stored on hard drives
2) Big Data = AI’s fossil fuel
3) Supercomputers + Big Data → AGI
4) AGI = “solves” the economy
3/
Two Insurmountable Flaws:
đźš« Flaw 1:
The economy isn’t a math puzzle to be *solved*.
It’s a *process*.
đźš« Flaw 2:
The most important economic knowledge is *uncodifiable*.
Even the most advanced AGI is constitutionally blind to it.
Two Insurmountable Flaws:
đźš« Flaw 1:
The economy isn’t a math puzzle to be *solved*.
It’s a *process*.
đźš« Flaw 2:
The most important economic knowledge is *uncodifiable*.
Even the most advanced AGI is constitutionally blind to it.
4/
A century before ChatGPT, economists warred over central planning v. markets.
Top down versus bottom up.
Early debates focused on human nature and incentives: who cleans the toilets?
But central planning's Achilles heel runs far deeper than mere bad incentives.
A century before ChatGPT, economists warred over central planning v. markets.
Top down versus bottom up.
Early debates focused on human nature and incentives: who cleans the toilets?
But central planning's Achilles heel runs far deeper than mere bad incentives.
5/
Central planners are *economically blind*—even under *utopian conditions*.
Central planning's fatal flaw isn't computational or technical.
It's *categorical*.
The economic knowledge that matters *most* is fundamentally *uncodifiable*.
Central planners are *economically blind*—even under *utopian conditions*.
Central planning's fatal flaw isn't computational or technical.
It's *categorical*.
The economic knowledge that matters *most* is fundamentally *uncodifiable*.
6/ This special economic knowledge is immune to quantification, storage, and algorithmic processing.
No AI—no matter how advanced—can access it or replicate it.
It only emerges when *humans* act under specific circumstances, in real time.
No AI—no matter how advanced—can access it or replicate it.
It only emerges when *humans* act under specific circumstances, in real time.
7/ đź’ˇKey insight:
The price system is the original Big Data-based AGI in the cloud.
It’s humanity’s most sophisticated information system and processor.
Even *in principle* no AGI can replicate its magic—at least not in service of *human* ends.
The price system is the original Big Data-based AGI in the cloud.
It’s humanity’s most sophisticated information system and processor.
Even *in principle* no AGI can replicate its magic—at least not in service of *human* ends.
8/ The essential knowledge required for economic calculation and planning isn’t just sitting out there—waiting to be scraped, processed, or optimized.
It has to be created—continually, in context—by humans.
It has to be created—continually, in context—by humans.
9/ This knowledge can’t be synthesized.
It can’t be automated.
It can’t be conjured into existence by decree.
It forms like a rare crystalline structure—only when temperature, pressure, and chemical conditions align perfectly.
It can’t be automated.
It can’t be conjured into existence by decree.
It forms like a rare crystalline structure—only when temperature, pressure, and chemical conditions align perfectly.
10/ A real price isn’t just a *number*.
It encodes *human* judgment—what a *real human* values in a particular time, place, and circumstance.
It encodes *human* judgment—what a *real human* values in a particular time, place, and circumstance.
11/ The essential information embedded in prices only emerges when it is generated by individual *humans* operating within the institutional context of a free market:
• Private property
• Sound money
• Voluntary exchange
• Private property
• Sound money
• Voluntary exchange
12/ A *genuine price* encodes information that no AI can access.
No AI—no matter how advanced—can force or simulate the conditions that make it emerge.
It has to be *lived*. By a human.
This is the epistemological line AGI cannot cross.
No AI—no matter how advanced—can force or simulate the conditions that make it emerge.
It has to be *lived*. By a human.
This is the epistemological line AGI cannot cross.
13/ Prices are *humanity's* most sophisticated coordination system.
They allow large-scale societal coordination across space and time.
They allow large-scale societal coordination across space and time.
14/
Prices tell us:
• What to make
• When and how to make it
• Which scarce inputs to use in making it
Remove their institutional scaffolding, and prices collapse into noise—whether in the USSR or tomorrow’s algorithmically-directed economies.
Prices tell us:
• What to make
• When and how to make it
• Which scarce inputs to use in making it
Remove their institutional scaffolding, and prices collapse into noise—whether in the USSR or tomorrow’s algorithmically-directed economies.
15/ Any discussion of “solving” the economy as if it were a math problem reveals a fundamental misunderstanding.
The market process and price system are “computationally irreducible”, as @stevenwolfram might say.
The market process and price system are “computationally irreducible”, as @stevenwolfram might say.
16/ Try to shortcut the market process, and you kill its magic.
Automate “prices” and you erase their meaning.
Automate “prices” and you erase their meaning.
17/ AGI *can* probably plan an economy—just not one for humans.
It can’t weigh meaning, values, or tradeoffs that matter to us.
An AI-planned economy might hum like a machine—
but it won’t feel like a civilization.
It can’t weigh meaning, values, or tradeoffs that matter to us.
An AI-planned economy might hum like a machine—
but it won’t feel like a civilization.
18/ In an AI techno-utopia with AGI, Quantum Supercomputers, Neuralink, and infinite Big Data—the AGI will know the *price* of *everything*.
And the *value* of *nothing*.
And the *value* of *nothing*.
19/ Without human prices, AGI can't compare tradeoffs.
Can’t allocate resources in a way that is meaningful for humans.
Can’t make economic decisions that serve *human* flourishing.
It can’t *plan*.
Because in an AI-planned economy, there’s nothing *real* left to plan *with*.
Can’t allocate resources in a way that is meaningful for humans.
Can’t make economic decisions that serve *human* flourishing.
It can’t *plan*.
Because in an AI-planned economy, there’s nothing *real* left to plan *with*.
20/ Even the smartest AGI will still be flying blind.
It’s not a technical limitation of data, compute or intelligence.
It’s an epistemological limit.
Even if we build AGI, it can never replicate or access the subjective, creative *human* process that gives prices *meaning*.
It’s not a technical limitation of data, compute or intelligence.
It’s an epistemological limit.
Even if we build AGI, it can never replicate or access the subjective, creative *human* process that gives prices *meaning*.
21/ This isn’t abstract theory.
It’s already happening, now.
Prices are being corrupted—by AI, by algorithmic code, by central banks, by passive ETF flows.
We call these distorted new signals #HyperPrices.
They *look* like prices.
But they’re artificial.
It’s already happening, now.
Prices are being corrupted—by AI, by algorithmic code, by central banks, by passive ETF flows.
We call these distorted new signals #HyperPrices.
They *look* like prices.
But they’re artificial.
22/ The imminent risk isn’t Skynet’s AGI building a Soviet-style command economy.
It’s that we’ve already wired our algorithms into Bloomberg terminals—streaming synthetic signals through real systems.
It’s that we’ve already wired our algorithms into Bloomberg terminals—streaming synthetic signals through real systems.
23/ We’re *already* simulating a market.
A synthetic price system built on HyperPrices.
Welcome to the Financial Matrix.
đź“– Read the full analysis at the pointer in our Twitter profile
A synthetic price system built on HyperPrices.
Welcome to the Financial Matrix.
đź“– Read the full analysis at the pointer in our Twitter profile
Read more here:
The Price Of Everything & Value Of Nothing
How Skynet Is Digitally Remastering History's Greatest Economic Debate bewaterltd.com/p/the-price-of…
The Price Of Everything & Value Of Nothing
How Skynet Is Digitally Remastering History's Greatest Economic Debate bewaterltd.com/p/the-price-of…
@threadreaderapp
unroll please
unroll please
• • •
Missing some Tweet in this thread? You can try to
force a refresh
　
