★★★★★ · 2025 #AI#tech#biology#philosophy

What Is Life?: Evolution as Computation

by Blaise Agüera y Arcas

The book in brief. This small book is the symbiogenesis argument of What Is Intelligence?, self-contained and consequential enough to have been published separately. Agüera y Arcas treats life as computation: a phase of matter that captures energy to maintain and copy its own structure. Complexity arises not through competition alone but through formerly separate entities merging into new cooperative wholes.

Key ideas

  • The universal constructor. In 1951 von Neumann showed that a self-reproducing system needs a constructor and a tape of instructions, first read and then copied blindly.
  • Life as a phase of matter. Living matter holds a dynamic stability, like a bicycle kept upright by pedalling.
  • Two engines of evolution. Selection explains refinement within a level, while symbiogenesis, after Lynn Margulis, explains the jumps between levels.
  • Compression. A genome is a compressed model of the environment its organism must survive in, and compression is always relative to that environment.
  • The next transition. The logic that produced mitochondria is now at work on the partnership between a brain and a large language model.

Why it matters to my work. My notes are in Reading What Is Intelligence? (1): Life Is Computation, and AI as the Next Symbiogenesis, where I add a connection the author does not make explicitly, between von Neumann’s blind copy step and Gödel’s limits on self-description. In On Thinking with (AI) Machines symbiogenesis lets me think of AI as something we are metabolising and not as an external tool. The genome as compressed model returns in Compressing a Causal Chain: What Has to Be True.