Discussion about this post

User's avatar
Richard Careaga's avatar

Technological progress can be compared to scientific progress as seen by Thomas Kuhn.

Most “normal” scientific work consists in the filling of gaps and exploring edge cases at the intersection of theory and fact. Eventually, the contradictions between theory and expanding understanding of reality forces a replacement theory—a paradigm shift that resolves the contradictions and starts a new cycle of discovery of new conflicts.

In 1968, the introductory textbook for historical geology I read as an undergrad had it the the movement of continents up and down explained all of what we can see by way of the arrangement of igneous, sedimentary and metamorphic rocks. The competing theory that continents moved laterally was rejected not only because there was no apparent mechanism to drive it, but because it was unnecessary to account for observations.

In 1972, I taught undergraduates from the next edition of the same text. Paradigm shift. Now, the notion that vertical movement of continents had explanatory power was restricted to a relatively minor role. The real action came from the lateral movement of continents and the larger plates beneath them. Surface and sea floor geology was now part of the same framework. Plate tectonics blew away the accumulated encrusting descriptions based on the old paradigm just as the Copernican Revolution blew away the circles-within-circles of the Ptolemaic elaborations necessary to reconcile astronomical observation in accordance with the geocentric view.

Most of what we think of as technological progress is the “D” in R&D. Exhausting the implications of an idea is necessarily a diminishing returns proposition. And, as a variations on a theme process is amenable to benefiting from systematic organization, it is a good candidate for AI augmentation. Perhaps there is even an iron law of conservation of required effort. As the human input per unit of improvement decreases with AI, that savings is shifted to productivity demand for human effort per unit of meta-structure.

The “R” arm is less amenable to AI augmentation. It requires recognition not only of the unsolved problems but identifying which of those are unsolvable under an existing framework and then imagining a framework under which they would be. in other words a paradigm shift. The work of that thinking cannot be outsourced to chips and code and management consists primarily of introducing smart people to the environment in the tradition of Bell Labs. Rather than being subject to a law of diminishing returns research in this sense is subject to a law of uncertain returns.

This process is what Alan Altshuler described as “disjoint incrementalism.” History proceeds in jumps and starts rather than monotonically or along any pre-specified curve. If so, it is cause for neither pessimism or optimism, only close observation.

Expand full comment
John Quiggin's avatar

Worth discussing the financial sector and austerity, I think

Expand full comment
8 more comments...

No posts