I appreciated this essay. I wonder what Hayek would make of AI? I think he would come to similar conclusions stated here.
My son's girlfriend asked me what I thought the difference was between AI and humans. I offhandedly said it is our errors. The more I thought about it, the words I had spoke with no real thought, were thoughtful. Western Civilization has been grounded in 'forgive them Father, for they know not what they do.' Sin is considered an error. But it is also the way for redemption and a deeper reflective life. Errors can also lead to scientific discoveries. How many times have we heard that an inventor intended to create one thing and did something completely off course? Will humanity remain without errors?
An interesting application for AI systems is in helping us to be more satisficing in our choices, which has been shown to increase happiness. Amazon's suggested search results are a good example of this working without agentic AI, in that the results are consistently good enough that I find little value in scrolling for more options. Obviously this indifference gives Amazon pricing power. One can imagine agentic AI handling the optimizing on the consumer side, allowing us to offload our cognitive load for inconsequential decisions so we can spend more time and energy engaged in the projects that are most meaningful to us.
The core claim sounds familiar: just as people once imagined that the internet would do away with gatekeepers, this piece assumes that AI agents will make coordination more abundant, freer, and more decentralized. But the internet did not remove intermediaries; it created new ones. Search engines, platforms, app stores, cloud providers, and payment rails became the new choke points.
AI agents may reduce some transaction costs, but that does not mean ordinary users suddenly gain real bargaining power. More likely, coordination simply gets repackaged through rankings, permissions, defaults, commissions, data access, and platform control. The question is not whether agents can "know us better." The question is who owns them, who pays them, who audits them, and whether users can leave.
So the optimistic version is plausible only in a weak sense: agents may help with search, matching, scheduling, and negotiation. The stronger claim — that they naturally produce a freer form of coordination — is essentially the same mistake early internet idealists made.
I appreciated this essay. I wonder what Hayek would make of AI? I think he would come to similar conclusions stated here.
My son's girlfriend asked me what I thought the difference was between AI and humans. I offhandedly said it is our errors. The more I thought about it, the words I had spoke with no real thought, were thoughtful. Western Civilization has been grounded in 'forgive them Father, for they know not what they do.' Sin is considered an error. But it is also the way for redemption and a deeper reflective life. Errors can also lead to scientific discoveries. How many times have we heard that an inventor intended to create one thing and did something completely off course? Will humanity remain without errors?
An interesting application for AI systems is in helping us to be more satisficing in our choices, which has been shown to increase happiness. Amazon's suggested search results are a good example of this working without agentic AI, in that the results are consistently good enough that I find little value in scrolling for more options. Obviously this indifference gives Amazon pricing power. One can imagine agentic AI handling the optimizing on the consumer side, allowing us to offload our cognitive load for inconsequential decisions so we can spend more time and energy engaged in the projects that are most meaningful to us.
The core claim sounds familiar: just as people once imagined that the internet would do away with gatekeepers, this piece assumes that AI agents will make coordination more abundant, freer, and more decentralized. But the internet did not remove intermediaries; it created new ones. Search engines, platforms, app stores, cloud providers, and payment rails became the new choke points.
AI agents may reduce some transaction costs, but that does not mean ordinary users suddenly gain real bargaining power. More likely, coordination simply gets repackaged through rankings, permissions, defaults, commissions, data access, and platform control. The question is not whether agents can "know us better." The question is who owns them, who pays them, who audits them, and whether users can leave.
So the optimistic version is plausible only in a weak sense: agents may help with search, matching, scheduling, and negotiation. The stronger claim — that they naturally produce a freer form of coordination — is essentially the same mistake early internet idealists made.