Theme 13 of 13 · 12 ideas

The book’s answer. It asks for a particular kind of expert, a particular kind of institution, and a longer time horizon than anyone would like.

The same ground, told as a story: Chapter 8: Preservation vs. Training, Chapter 9: The Integration Problem, Chapter 12: Infrastructure for Integration, Bridge Nodes and Versatile Expertise.

The ideas

Each one is a short page. Walk them in this order, or start with whichever catches your eye.

  1. Transferable vs. specialized preconditions (open question): Carrying knowledge between fields takes a deep specialist who has learned to step outside their own field’s habits, and general thinking skills alone fall short.
  2. The curse of expertise: Knowing one field very deeply can make it harder to see, or take seriously, anything that does not fit that field’s way of thinking.
  3. Versatile expertise: The people who can carry hard ideas between fields are deep specialists who also know their own way of seeing is only one of several.
  4. Preservation vs. training (paired institutional functions): Keeping hard knowledge alive takes two jobs at once: storing the full original somewhere, and continually teaching people how to read it.
  5. Differential decay rates (preservation vs. training): A neglected library can sit intact for decades, but skilled people must be trained fresh every generation, so teaching fails first.
  6. Curation layer: Let anyone contribute, then use open, layered review to sort the good from the bad, as Wikipedia does.
  7. Funding decoupling from attention markets: An institution that does slow, careful work has to be paid in some way that does not depend on winning people’s attention.
  8. Survivable polarization (design for partial trust): Build institutions that still work when some people distrust them, by letting doubters check the evidence and reasoning for themselves.
  9. Common-law courts as a curation institution (example): Common-law courts have kept hard knowledge alive for centuries by training judges for decades, writing everything down, and being paid for by the public.
  10. Stack Overflow: curation without funding decoupling (example): A programming Q&A site with strong quality control lost its audience when AI began answering the same questions, and it had no funding to fall back on.
  11. Survival, not victory (integration institutions): Careful institutions cannot beat the attention economy, so the goal is to keep doing their work alongside it without being worn down.
  12. Work of generations (institutional timescale): Institutions that keep hard knowledge alive take decades or centuries to build, so rebuilding them is a job for generations.

← AI in the chain · All themes