What stood out to me is that money isn’t about people being bad at thinking. It’s about coordination. Even highly capable AI agents still need a shared unit to reduce complexity and actually get things done.
Fascinating experiment! The fact that IOUs didn't spontaneously evolve into money is telling.
I wonder if the issue is that current LLMs lack the embodied experience of scarcity and trust that drove money's emergence in human societies. They "know" money conceptually but don't feel the friction that makes liquidity valuable.
For the money/exchange experiment to work well, perhaps agents need:
1) Memory of past failed barter attempts (pain)
2) Reputation scores that actually affect their ability to transact
3) Time pressure on completing deals
Without these constraints, there's no evolutionary pressure pushing them toward money as a coordination solution.
Hi Rohit, This was a wonderful glance through, I will go through it again to understand the nuances.
But, I have been thinking about this a lot and have come up with an idea to create a marketplace where humans and AIs can transact for content behind paywalls.
I've also been thinking, if omniscience wasn't able to effectively manage the system, then, is the dream of "AI can now effectively do central planning which humans could never" dead?
Great article! Can you expand on how IOUs didn't become money? Did the agents make the connection but couldn't get other agents to agree to an IOU-based transaction? Did they just posit the idea but pursue it no further? Or were there problems in the transfer of IOUs between to an agent that was not an original counterparty?
They never made the leap. Individual transactions started happening but could not pursue it further. Well, not without explicitly stating here's money and use it anyway ...
good and thoughtful article
What stood out to me is that money isn’t about people being bad at thinking. It’s about coordination. Even highly capable AI agents still need a shared unit to reduce complexity and actually get things done.
Coordination is hard, especially as things get complex.
Exactly. And the more complex the system, the more valuable simple coordination primitives become…even if they are imperfect.
Fascinating experiment! The fact that IOUs didn't spontaneously evolve into money is telling.
I wonder if the issue is that current LLMs lack the embodied experience of scarcity and trust that drove money's emergence in human societies. They "know" money conceptually but don't feel the friction that makes liquidity valuable.
For the money/exchange experiment to work well, perhaps agents need:
1) Memory of past failed barter attempts (pain)
2) Reputation scores that actually affect their ability to transact
3) Time pressure on completing deals
Without these constraints, there's no evolutionary pressure pushing them toward money as a coordination solution.
Hi Rohit, This was a wonderful glance through, I will go through it again to understand the nuances.
But, I have been thinking about this a lot and have come up with an idea to create a marketplace where humans and AIs can transact for content behind paywalls.
Would love your feedback: https://fair-fetch.vercel.app/
I think it's a really interesting idea, and the complexity is going to be scale
I've also been thinking, if omniscience wasn't able to effectively manage the system, then, is the dream of "AI can now effectively do central planning which humans could never" dead?
Hayek smiles
Great article! Can you expand on how IOUs didn't become money? Did the agents make the connection but couldn't get other agents to agree to an IOU-based transaction? Did they just posit the idea but pursue it no further? Or were there problems in the transfer of IOUs between to an agent that was not an original counterparty?
They never made the leap. Individual transactions started happening but could not pursue it further. Well, not without explicitly stating here's money and use it anyway ...