What Will Become Scarce in the AGI Era? Why Machines May Exclude Human Workers

1. From Partial Automation to Full‑Supply‑Chain Transformation

As artificial‑intelligence capabilities advance, the interview on the Dwarkesh Podcast between Alex Imas, AGI economics lead at Google DeepMind, and Phil Trammell, economics lead at Epoch, examines how a fully automated supply chain could reshape global macro‑economic factors.

2. Defining the AGI Era as a Qualitative Shift

Trammell defines the AGI era as a complete, qualitative transformation of the supply chain. Unlike the industrial‑revolution‑era “traditional automation” that replaces isolated tasks, a product whose entire chain—from raw material to delivery—is taken over by AI systems and robots requires no human involvement. In that scenario the "network‑adjusted capital share" of the product reaches 100%.

3. Implications for Traditional Economic Modeling

Imas argues that because AGI systematically removes both physical and cognitive human labor, conventional linear financial planning and incremental forecasting become invalid. He suggests adopting an extreme premise (labor share set to zero) and using prediction markets to work backwards from target outcomes, thereby re‑examining wealth distribution under expanding compute capital.

4. Lessons from Classical Economics

Imas cites David Ricardo’s industrial‑revolution analysis, noting Ricardo fell into the "labor‑block fallacy" by failing to anticipate that partial automation would free purchasing power and create new employment, keeping the macro labor share above 60%.

5. The Last Scarcity: Relationship‑Based Departments

The authors identify "relationship‑based departments"—areas where humans remain in the loop, such as psychotherapy or senior mentorship—as the final scarce resource. Behavioral‑economics experiments show consumers are willing to pay significantly more for artworks explicitly created by humans than for comparable AI‑generated pieces, indicating a premium on authentic human identity in a low‑cost compute future.

6. Alternative Perspective on Capital Concentration

Trammell introduces a thought experiment drawn from ancient Mongolian economics, arguing that AGI‑driven compute assets exhibit anomalous behavior. While Moore’s law traditionally halves the marginal value of computation, top‑tier AI accelerator rentals (e.g., NVIDIA H100) have risen in price, confirming that global wealth is increasingly concentrating in compute clusters.

7. Risks of Unbounded Capital Share Expansion

As AGI continuously unlocks new automated product categories, humans will keep investing wealth into machine outputs. If the diversification speed of compute applications outpaces the growth of consumption for relationship‑based services, the macro capital share can expand without bound, reducing the relative economic weight of human‑centric sectors.

8. The "Messy Middle" Transition and Political‑Economic Crises

After outlining the end‑state of full supply‑chain automation, the discussion turns to the transitional period—dubbed the "Messy Middle"—where AI gradually replaces cognitive labor. The economists use O‑ring theory to reveal a reliability paradox that firms face when deploying AI in real‑world environments, highlighting potential political‑economic crises during this gray zone.