Some interesting insights from a recent Dwarkesh Patel podcast - particularly in light of yet another Big Four scandal here in Australia, where client trust and confidentiality were compromised.

The most clarifying question in AI right now is not, “What can models do?” It’s, “What will remain scarce?”

Economists Alex Imas and Phil Trammell framed it well in the Patel interview: if intelligence becomes abundant, the remaining scarce inputs will determine who captures value. Their answer: attention, trust, embodied presence, physical constraints, and things that can only scale at a human pace. The number of ballerinas, for example, doesn’t increase simply because robots can dance.

For strategists and operators, this is a far more useful lens than debating model capabilities. The real question is: what does your business provide that cannot be replicated by running more inference?

If the answer is “not much”, that’s a genuine risk.

If the answer is relationships, institutional trust, regulatory standing, physical deployment, or other forms of hard-earned credibility, your position may be more durable than it appears.

The conversation also highlighted something underappreciated: AGI may accelerate capability, but it is likely to concentrate value. Good judgement at the application layer - deciding which problems to apply AI to, and how to evaluate the outputs - does not become less important. It becomes more important.

What are you treating as a durable moat that may, in fact, be exposed?

#strategy #AI #economics

https://lnkd.in/gAfEcyzn