AI Safety

Metaphilosophy II essay: empirical flywheels drive conceptual progress

A new essay reveals how sextants, breeding experiments, and X-rays update our highest-level concepts.

Deep Dive

The essay 'Metaphilosophy II' opens by defining philosophy as updating the highest-level concepts of the mind. It distinguishes two complementary methods: first, drawing out connections among existing concepts (the traditional approach, showing contradictions or unifying seemingly different ideas); second, seeking experiences and knowledge especially likely to help update those high-level concepts. The author argues that choosing what to learn cannot be guided by a yet-higher concept—since that's what we're trying to update—so a pragmatic strategy is to study areas where we are maximally uncertain, but chosen based on their proximity to our current notions of reality and goodness.

This leads to the essay's core insight: the 'empirical flywheel.' Defined as a set of well-defined procedures and technologies enabling precise questions about physical or mathematical reality, flywheels create virtuous feedback loops between conceptual progress and empirical/calculational work. Historical examples include Tycho Brahe's sextant measurements that enabled Kepler's laws, Mendelian breeding experiments in biology, and X-ray diffraction experiments that revealed DNA's structure. The author argues that science and mathematics serve as natural flywheels for updating our highest-level concepts, making them a powerful strategy for philosophical progress in modernity.

Key Points
  • The essay defines two philosophical methods: conceptual connection-drawing and deliberately learning new things to update high-level concepts
  • Empirical flywheels are well-defined procedures like sextant astronomy or X-ray diffraction that generate precise data and drive conceptual breakthroughs
  • Virtuous feedback loops between flywheels and conceptual progress explain modernity's explosion in scientific and mathematical knowledge

Why It Matters

Understanding empirical flywheels can guide research strategies and help AI developers design systems that accelerate conceptual progress.

📬 Get the top 10 AI stories daily