TB: AI and Research Papers (Kling)

Core Thesis

The unit of scientific knowledge is the claim, not the paper. AI should enable a shift from paper-centric to claim-centric communication: a researcher asking "what do we know about X" should receive a structured, confidence-weighted answer — not a list of PDFs.

Key Takeaways

The wrong unit

  • The academic paper bundles together claims, evidence, methods, and argument — a useful package for communication between specialists, but a poor unit for knowledge retrieval.
  • AI makes it possible to unbundle: extract the claims, weight them by evidence quality, and synthesize across papers rather than listing them.

The implication

  • "A researcher asking 'what do we know about X' should get a structured confidence-weighted answer, not a list of PDFs to read."
  • This reframes what AI assistants should actually do for scientists — not summarize papers, but map the claim-space.

Mental Models

  • The Map is Not the Territory — the paper is the map; the claim is closer to the territory
  • Inversion — inverting from "what papers exist?" to "what do we know?" transforms the research task

See also

Source

Arnold Kling, Substack, 2026 Original article