Recursive technology, non-recursive adoption, and Schumpeterian displacement

Tyler pointed to this piece, from Citadel Securities, making the point Tyler's made before: AI is growing more powerful by leaps and bounds, but people and institutions are slow to adopt new things. 

Diffusion and bottlenecks within institutions will prevent an AI-powered growth explosion and rapid displacement. 

Recursive Technology ≠ Recursive Adoption

The current debate around artificial intelligence conflates the recursive potential of the technology with expectations of recursive economic deployment. In other words, because AI systems can improve themselves or accelerate their own capabilities, commentators are extrapolating a future in which automation and productivity compound indefinitely at exponential rates. Technological diffusion has historically followed an S-curve. Early adoption is slow and expensive. Growth accelerates as costs fall, and complementary infrastructure develops. Eventually, saturation sets in, and the marginal adopter is less productive or less profitable which causes growth to decelerate.

Despite this – markets often extrapolate the acceleration phase linearly but history implies pace of adoption plateaus as organizational integration is costly, regulation emerges and diminishing marginal returns exist in economic deployment. The risk of displacement declines with a slower pace of adoption.

I'm happy to agree with all that. 

And I think it was Josh Gans who, a few months ago, pointed to very wide within-industry dispersion of firm-level productivity as suggestive. Highly productive and far-less-productive firms coexist in single industries.

But if the productivity gap is large enough between firms that really get AI and those that don't, wouldn't we expect the laggards to be winnowed out? If old institutions are slow to move, but new upstarts can leapfrog them, you could have stats showing relatively small proportions of workers relying heavily on AI, as Citadel currently shows. But weighting that by market share could be different. 

I dunno. When Aella can quickly roll-out dataviz with Claude that beats anything Stats NZ has ever put up, it feels like there's potential for tiny firms to quickly eat large lunches.