Life is a spontaneously auto-catalytic process or system that decreases local entropy, subject to the zeroeth law of thermodynamics.

- Nick Porcino, August 2008

Corollary: A machine may be spontaneously auto-catalytic, but is only a machine if it increases local entropy.

Introduction

I had several days with hours at a stretch to think, and the definition for life above was what I got out of it. I like this definition, it's clean, concise, and doesn't have any of the usual sort of wobbly pitfalls of the usual sort of definition. It also doesn't reach for any arm-wavey consumption of intangibles such as negentropy, and doesn't need a table of exceptions or justifications about why one thing can be considered life and another can't. It might also help us make a call when aliens show up, or when a computer gets a bit too smart.

Overturning Convention

There is a nice article at Wikipedia describing the conventional definition of life, and lists some counter-examples that break convention. Life is often defined as an ad hoc melange of rules:

• homeostasis (regulation of internal state, something robots can do)
• organization (limits life to cellular organisms - virii are not life by convention)
• metabolism (consumption of energy to create cellular components)
• growth (maintaining a higher rate of synthesis than catabolism)
• adaptation (some simple organisms have remained largely the same for millions of years - are they therefore not alive?)
• response to stimuli
• reproduction (are worker bees not alive?)

One can fiddle away on these rules trying to refine them and catch border cases, but it never really pans out.

The systems theory definition of life states that live things self-organize and self-reproduce. This is closer to what I am after, but not strong enough to make predictions, or to address faults and ambiguities within the ad hoc conventional definition.

Other definitions are organized around agency, replication, or enzymes. All of these are narrowly focused around particular interpretations specific to particular fields of study.

Erwin Schrödinger wrestled with the question and came close to what I think is the right definition when he suggested that living systems import and store negative entropy or negentropy. Where this fails for me is that negentropy is an information theory statistic used as a measure of local normality; in my definition I am using entropy as a measure of the ability of a given chunk of space-time to do work. So, just as quantum mechanics describes physical processes without revealing the reality of the physicality, the storage of negative entropy describes what life is getting up to, but it does not tell what is going on. To put it formally, statistics and chance are purely epistemological concepts without ontological import.

Reading my new Definition

Spontaneously auto-catalytic system or process

Life happens on its own; it bootstraps itself; when the time is right, it picks up and goes. A seed in the presence of moisture begins to manufacture a plant. A quickened ova divides as factories in the cell's nucleus begin to run. This is not sufficient for a definition of life. it encompasses things that we unambiguously do not consider alive such as fire, a wound-up wall clock, or reaction-diffusion systems.

Decreasing local entropy

There are a number of ways to look at this; the easiest is to imagine winding a clock-spring tighter. When that spring is wound, the potential is there to drive the clock. One can imagine that life winds the spring constantly through some kind of process of conversion of environmental energy to potential in the spring. To pick a life-form excluded by convention, virii wind their spring by organizing matter into self-replicating space-time geometries. Fire is excluded from being considered alive, because it unwinds the springs; entropy is increased by fire. By itself, the clock is not alive, it can only wind the spring down and is therefore a machine.

I suggest the words "decreasing local entropy" to help with conceptualization of the definition; one can imagine points of life distorting an entropic landscape similarly to the visualizations of black holes warping space-time.

Surprising Inductees to the Hall of Life

My definition welcomes back the worker bee, and the virus. My definition also allows bacteria that can only survive in host cells. It will make it easier to classify alien life should it present itself some place like Mars in some difficult to conventionally pigeonhole fashion.

Some things that we describe in terms of living and dying are reminiscent of life, but are not life - companies, cities, and nations do not auto-catalyze although it sometimes appears that they do. I'm including this note because it is interesting to debate.

It would be possible for robots or computers, or even the Internet to be alive, although present day versions are not spontaneously auto-catalytic, and do not inherently decrease local entropy since they must be manufactured, consume energy, and do not effectively wind springs in return.