AI Safety

LessWrong's Fun Formula: A Mathematical Framework for Transhumanist Futures

Three operations on a 'fun field' redefine how we measure a good future.

Deep Dive

In a recent LessWrong post, Ihor Kendiukhov presents 'A Formula for Fun,' a semi-formal attempt to describe preferences for the transhuman future by modeling good experiences as a continuous 'fun field' spread across all minds, places, and moments. The field peaks where flourishing is intense and drops to zero elsewhere. Kendiukhov then identifies three distinct mathematical operations that could be applied to this field, representing different ethical intuitions: total fun (summation over all beings and time), variety of fun (counting distinct kinds of good experiences regardless of repetition), and rate of change of fun (how fast the field grows).

Each operation carries well-known philosophical baggage: total fun leads to the repugnant conclusion where a vast number of barely happy beings outweighs a smaller radiant population; variety of fun runs into the problem of granularity—how finely we slice experiences determines the result; and rate of change prioritizes acceleration over absolute level, risking instability. Kendiukhov suggests a unified formula that combines these as special cases, aiming to capture what's compelling in each while avoiding their failure modes. The work is less a prescription and more an exercise in making implicit preferences explicit, contributing to alignment and fun theory discussions.

Key Points
  • Proposes a 'fun field' representing flourishing density across time, minds, and space.
  • Three aggregation operations: total fun (sum), variety (distinct kinds), and rate of change (growth).
  • Each operation has known pitfalls: repugnant conclusion, slicing granularity, and acceleration bias.

Why It Matters

Provides alignment researchers a formal toolkit to debate which future metrics to optimize for.