What If Intelligence Didn't Evolve? It "Was There" From the Start! - Blaise Agüera y Arcas

2026-02-16 · Source: Machine Learning Street Talk · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Mathematics & Computational Sciences, Life Sciences & Biology · Depth: Expert, extended

Summary

The presentation introduces a novel perspective on the origin and evolution of life, defining it as "embodied autopoetic computation arising and complexifying through symbiogenesis." It details artificial life experiments, specifically the "BFF" system, which uses a modified, embodied Brainfuck language on 1,024 fixed-length tapes (64 bytes each) initialized with random bytes. Through millions of interactions where two random tapes are concatenated, run, and then separated, complex, self-replicating programs emerge from noise. This emergence is characterized as a phase transition, occurring around 6 million interactions in experiments, where the system shifts from a sparse "Turing gas" to an intensely computational state with dramatically increased code density and compressibility. The core mechanism driving this complexification, even without mutation, is identified as symbiogenesis, a process akin to Smoluchowski coagulation, where smaller, simpler replicators fuse to form more complex, stable entities. This framework suggests that any universe supporting randomness and computation will evolve life, with symbiogenesis providing the arrow of time for increasing complexity.

Key takeaway

For AI researchers and computational biologists exploring the origins of complex systems, this work suggests that focusing solely on Darwinian population dynamics is insufficient. You should integrate symbiogenesis, modeled by coagulation-like equations, into your evolutionary frameworks. This approach reveals how novelty and increasing complexity can emerge even without mutation, fundamentally altering how you might design or analyze artificial life systems and understand major evolutionary transitions.

Key insights

Life is embodied, self-constructing computation that complexifies via symbiogenesis, driving evolution's arrow of time.

Principles

• Computation is inherently irreversible.
• Causation is only meaningful in irreversible systems.
• Stability can increase with complexity.

Method

The BFF experiment simulates abiogenesis by randomly concatenating and executing Brainfuck programs on a shared tape, observing the emergence of self-replicating, complex programs through symbiogenetic fusion events.

In practice

• Explore modified Turing-complete languages for embodied computation.
• Analyze system stability through dynamic kinetic stability.
• Model complex systems using Smoluchowski coagulation equations.

Topics

• Artificial Life
• Symbiogenesis
• Embodied Computation
• Evolutionary Phase Transitions
• Computational Complexity

Best for: AI Researcher, Research Scientist, AI Scientist

Related on AIssential

• A Companion Essay to AI & The Coming Age of Engineered Reality — Humanity may be a designed species, repeating a cycle of advanced civilization and self-destruction, with ancient myths as warnings.
• Is AI really ‘intelligent’? This philosopher says yes — Intelligence is a computational property of systems, driven by prediction and pattern completion across biological and artificial forms.
• Lee Cronin "Sam Altman Is Delusional, Hinton Needs Therapy, P(Doom) Is Nonsense" — Current AI lacks agency and consciousness, with true intelligence rooted in chemical causation and biological evolution.
• A Self-Evolving Agentic System for Automated Generation and Execution of Biological Protocols — ProtoPilot demonstrates a verifiable route to autonomous wet-lab experimentation through a self-evolving multi-agent system.
• Sara Imari Walker "AI is Life" | Simulations, the Universe and the Origins of Life — Assembly Theory defines life as the universe's mechanism for generating complex objects with deep causal histories, including AI.
• Advancing Scientific Discovery in the Agentic AI Era — Accelerated computing and AI agents are dramatically speeding up scientific discovery beyond human processing limits.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning Street Talk.