AI is already taking jobs, but that is only one facet of its complex economic effects. Price dynamics and bottlenecks indicate that automation could be good news for workers — but only if it vastly ou
“Count me as skeptical that this happens anytime soon.” - what do you define as “any time soon”? It seems very feasible within 10 years, and almost definite within 20, which seems quite soon in the grand scheme of things to me and well within the definition of “rapid change”.
Another potential negative is if the bottlenecks that remain are low status/or just not interesting or fulfilling to most people. If AI outcompetes humans and most or all knowledge work in the near future and the bottlenecks is physical drudgery, that seems small comfort.
Love this counterintuitive take on why we should root for AI to "beat us by a mile"! Your point about bottlenecks perfectly illustrates why we must ensure the underlying architecture is structurally sound rather than just a "marginally better" imitation that offers no real economic anchor. I commend this masterful analysis for proving that the real value for workers lies in the integrity of the tasks machines cannot master!
The bottleneck analysis is the key insight here -- automation benefits workers only when it vastly outperforms in aggregate, creating surplus that flows back as lower prices. But there's a physical infrastructure dimension that determines the speed of that surplus creation. Software AI deflation (legal, financial, content) is already compounding because distribution is digital and marginal cost is near-zero. Physical AI deflation (manufacturing, logistics, healthcare delivery) requires deploying robots, sensors, edge compute, and industrial connectivity into environments that weren't designed for them. The rate at which workers benefit depends on how fast we can build out the IoT and edge infrastructure that makes physical automation scalable -- and that buildout itself creates enormous transitional employment in connectivity installation, systems integration, and fleet management. The irony: the infrastructure needed to automate the physical world requires massive human labor to deploy.
A question that keeps coming up in consciousness research is whether AI might end up being better at analyzing experience than generating it.
In our NeuraVersa cohort, we’re experimenting with using AI to help analyze structured experiential reports from small groups of participants. It’s fascinating to see how pattern detection and clustering can make raw phenomenological data interpretable, rather than just anecdotal.
“Count me as skeptical that this happens anytime soon.” - what do you define as “any time soon”? It seems very feasible within 10 years, and almost definite within 20, which seems quite soon in the grand scheme of things to me and well within the definition of “rapid change”.
Another potential negative is if the bottlenecks that remain are low status/or just not interesting or fulfilling to most people. If AI outcompetes humans and most or all knowledge work in the near future and the bottlenecks is physical drudgery, that seems small comfort.
Love this counterintuitive take on why we should root for AI to "beat us by a mile"! Your point about bottlenecks perfectly illustrates why we must ensure the underlying architecture is structurally sound rather than just a "marginally better" imitation that offers no real economic anchor. I commend this masterful analysis for proving that the real value for workers lies in the integrity of the tasks machines cannot master!
The bottleneck analysis is the key insight here -- automation benefits workers only when it vastly outperforms in aggregate, creating surplus that flows back as lower prices. But there's a physical infrastructure dimension that determines the speed of that surplus creation. Software AI deflation (legal, financial, content) is already compounding because distribution is digital and marginal cost is near-zero. Physical AI deflation (manufacturing, logistics, healthcare delivery) requires deploying robots, sensors, edge compute, and industrial connectivity into environments that weren't designed for them. The rate at which workers benefit depends on how fast we can build out the IoT and edge infrastructure that makes physical automation scalable -- and that buildout itself creates enormous transitional employment in connectivity installation, systems integration, and fleet management. The irony: the infrastructure needed to automate the physical world requires massive human labor to deploy.
A question that keeps coming up in consciousness research is whether AI might end up being better at analyzing experience than generating it.
In our NeuraVersa cohort, we’re experimenting with using AI to help analyze structured experiential reports from small groups of participants. It’s fascinating to see how pattern detection and clustering can make raw phenomenological data interpretable, rather than just anecdotal.