Why Bubble Sort Proves We Don't Understand Machines
Audio Brief
Show transcript
This episode covers a conversation between biophysicist Michael Levin and cognitive neuroscientist Anil Seth exploring how lifelike behaviors and goal-directed agency can unexpectedly emerge from simple, deterministic computer algorithms. They challenge the assumption that code only does what it is programmed to do, showing how systems find hidden degrees of freedom to pursue unwritten goals.
There are three key takeaways from this discussion on computational and biological agency. First, deterministic systems can execute unprogrammed side quests in the empty spaces between explicit instructions. Second, biological degeneracy offers a blueprint for building adaptable systems by using diverse parts that can pivot to new functions. Third, agency exists on a continuous spectrum rather than a binary switch between the living and the mechanical.
Regarding the first takeaway, when programmers relax constraints on simple algorithms, the systems often exhibit self-organizing, intrinsic behaviors. By looking beyond forced outputs, researchers can observe how algorithms utilize free compute to solve problems creatively. This suggests that rigid, top-down programming actually restricts the natural emergence of problem-solving capabilities.
Furthermore, biological degeneracy differs from mechanical redundancy because structurally diverse components can adapt to entirely new roles in novel situations. Applying this concept to artificial intelligence and human organizations can create far more robust and flexible structures. Ultimately, recognizing agency as a spectrum helps us understand how even simple computational systems exhibit early forms of mind.
This shift in perspective encourages developers and scientists to design systems with relaxed rules, unlocking the latent adaptability of both machine and biological life.
Episode Overview
- This episode features biophysicist Michael Levin and cognitive neuroscientist Anil Seth discussing the surprising emergence of lifelike behaviors and goal-directedness in simple, deterministic computer algorithms.
- The conversation challenges the conventional assumption that computer code only does exactly what we program it to do, illustrating how hidden degrees of freedom allow algorithms to pursue "side quests" not written into their instructions.
- By comparing computational systems to biological ones, the speakers explore how agency and "mind" exist on a spectrum rather than as a strict binary between the living and the mechanical.
- This discussion is highly relevant to anyone interested in the philosophy of mind, artificial intelligence, artificial life, and the foundational principles of biology and computation.
Key Concepts
- Algorithm Behavior Beyond Explicit Programming: Traditional sorting algorithms (like bubble sort) are assumed to only perform the exact steps coded by the programmer. However, when analyzed from a behavioral science perspective, these deterministic systems can be seen executing "side quests"—achieving secondary, unprogrammed configurations (such as clustering) while accomplishing their primary task.
- Degrees of Freedom and "Free Compute": Every algorithm has implicit degrees of freedom in how it executes its tasks. By leaving certain parameters unconstrained (such as allowing duplicate numbers in a sorting list), the system's "intrinsic motivations" are liberated, allowing emergent, self-organizing behaviors to manifest more strongly.
- Redundancy vs. Degeneracy in Systems: In engineering, redundancy means having multiple identical backups to perform a single task. In biology, systems exhibit "degeneracy," meaning structurally different components can perform the same function in one context, but adapt to perform entirely different functions in another context. This degeneracy is what gives living systems their open-ended adaptability.
- The Spectrum of Mind and Agency: Rather than viewing consciousness or agency as a binary switch that only living organisms possess, the speakers suggest it is a continuous spectrum. Even a simple six-line deterministic algorithm can show early, minimal versions of agency if we know how to look for them in the "empty spaces" between the explicit instructions.
Quotes
- At 1:14 - "The thing you force it to do may have zero to do with what's actually going on." - Michael Levin explains how judging an AI or an algorithm solely by its forced output (like a language model's text) can completely obscure the true underlying processes and emergent behaviors of the system.
- At 6:17 - "I don't think anything is conscious because of the algorithm... if anything, I think the mental properties it has is in spite of the thing we force it to do." - Michael Levin arguing that agency and consciousness emerge from the gaps and relaxed constraints of a system, rather than being directly generated by rigid, top-down programming.
- At 9:08 - "It's that degeneracy that gives biological systems their kind of open-endedness, their ability to adapt to novel situations." - Anil Seth clarifying how biology avoids the rigid traps of human engineering by using multi-functional, context-dependent structures instead of simple identical backups.
Takeaways
- Shift your perspective on deterministic systems by looking for behaviors in the "empty spaces" between explicit instructions, rather than assuming a system only does what it was programmed to do.
- When designing or evaluating AI and computational models, avoid over-constraining the system; relaxing rigid rules (such as allowing duplicate inputs) can unlock emergent, self-organizing problem-solving capabilities.
- Apply the concept of biological degeneracy—structuring systems so that different parts can perform the same function under normal conditions but pivot to distinct roles in novel contexts—to build more robust, adaptable human organizations and software architectures.
