Life: An Interdisciplinary Inquiry

• Life is an interdisciplinary concept defined by adaptive, self-organizing, and information-processing systems across biological, thermodynamic, and computational contexts.
• Research employs thermodynamic frameworks, computational automata, and biosignature detection methods to elucidate life’s emergent properties.
• Diverse theories of abiogenesis, artificial life, and astrobiology highlight the challenges in distinguishing living from non-living systems.

Life is treated in contemporary research as an interdisciplinary enigma, a physicochemical continuum, a self-regulating process, and a nonequilibrium information-processing phenomenon. Across biology, thermodynamics, astrobiology, artificial life, and formal computational systems, no single universally accepted definition governs all usage. Some authors argue that there is no comprehensive definition of living beings and no specific temporal boundary marking the onset of life, while others propose mechanistic or thermodynamic criteria centered on directed work, open-system growth, self-assembly, information storage, and adaptation (Azar, 2022, Pierce, 2020, Jones, 14 Jun 2026).

1. Definitional landscape

A persistent starting point is the cluster of familiar biological traits: growth, reproduction, responsiveness to stimuli, adaptation, homeostasis, metabolism, self-organization, compartmentalization, and information processing. One line of argument holds that none of these alone, and no finite conjunction of them, uniquely defines life, because each can appear at least partially in non-living or borderline systems. On this view, living organisms do not possess unique characteristics that can completely set them apart from non-living entities, and the transition from non-living to living matter is continuous rather than sharply bounded (Azar, 2022).

That position is formalized by introducing two sets, $A$ for living creatures and $B$ for non-living creatures, together with a finite family of candidate properties $q_1(x), q_2(x), \ldots, q_n(x)$. The same paper argues that any attempt to define $A$ by a subset $Q$ of these properties is circularly dependent on how the sample of living beings is chosen. Its continuous-transition argument uses a function

$f:[a,b]\to[0,1], \qquad f(a)=0,\quad f(b)=1,$

and claims that if $f$ is continuous, any precise boundary between non-life and life becomes arbitrary (Azar, 2022).

A contrasting approach seeks a mechanistic definition rather than a checklist of emergent traits. One proposal states: $$\text{Life is a self-regulating process whereby matter undergoes cyclic, uniplanar conformation state changes that convert thermodynamic disequilibria into directed motion, performing work that locally reduces entropy.}$$ A related object-level definition describes a living thing as a structure comprising, at least in part, an autonomous network of units exploiting thermodynamic gradients to drive uniplanar conformation state changes that perform work. In this framework, growth, self-replication, homeostasis, metabolism, evolution, and agency are downstream consequences rather than the underlying cause of life (Pierce, 2020).

A broader conceptual review describes life as self-perpetuating, evolving, exquisitely adaptable, and in specific instances, capable of bearing consciousness. It also distinguishes biological life, intelligence, and self-awareness rather than treating them as identical categories. A common misconception addressed in this literature is that reproduction or evolution alone settles the question; the mechanistic account explicitly argues that evolution and heredity occur across generations and cannot by themselves explain the instantaneous state of being alive in a cell at a given moment (Jiang et al., 2023, Pierce, 2020).

2. Thermodynamics, information, and adaptation

A major family of theories treats life as an open thermodynamic system maintained far from equilibrium. One explicit criterion requires two formal conditions: first, a growing open thermodynamic system out of equilibrium; second, a system that performs synthesis, self-assembly, and accumulation processes. This framework maps canonical biological properties onto thermodynamic analogues: ordered cell structure to a thermodynamic system, reproduction to division into subunits, growth and development to mass and energy accumulation, response to the environment to Le Chatelier’s principle, homeostasis to dynamic equilibrium, and evolutionary adaptation to successful adaptation using information (Popovic, 2018).

Within the same thermodynamic tradition, life is framed as a process that maintains local order while exporting entropy to its surroundings. Schrödinger’s emphasis on entropy export and stable hereditary information is extended by a recent proposal of six postulates for adaptable life: the existence of an entropy source, longevity of information, fast response to environmental change, repeatable operation, energetic efficiency, and networks of multiple interacting switches. The same work treats natural selection as a statistical process that causes information about the environment to accumulate in a population of replicating systems, thereby integrating replication, variation, differential reproduction, nonequilibrium thermodynamics, and information processing (Jones, 14 Jun 2026).

An information-theoretic formulation pushes the argument across scales from molecules to ecosystems. In that view, life is information processing, memory is maintained by both molecular states and ecological states as well as nucleic acid coding, and the overall function of that processing is to perpetuate itself. The paper distinguishes total information from functional information through

$I_{tot} = I_F + I_R,$

where $I_F$ is functional information and $B$0 is random information. Functional information is the part capable of causing persistent change in the broader system. In the paper’s computer analogy, life is both the data and the program, and its biochemical structure is the way the information is embodied (Farnsworth et al., 2012).

These thermodynamic and informational accounts converge on several shared motifs. Living systems exploit thermodynamic disequilibria, store information in metastable substrates, and use cyclic switching or contextual organization to perform work repeatedly. They differ, however, in emphasis: one centers molecular heat-engine-like conformational cycles, another centers open-system growth and self-assembly, and another centers multiscale functional information (Pierce, 2020, Popovic, 2018, Farnsworth et al., 2012).

3. Origins, abiogenesis, and cosmic timescales

Origin-of-life research in this corpus is notably heterogeneous. One hypothesis extrapolates the genetic complexity of organisms backward in time and argues that functional, nonredundant genome complexity grew approximately exponentially, doubling about every $B$1 million years. Linear regression on a logarithmic scale extrapolated to one base pair yields an origin time of $B$2 billion years ago, implying that life began before the Earth was formed and that Earth was seeded by panspermia rather than being the site of first origin (Sharov et al., 2013).

A very different proposal, advanced by Abraham Loeb, links early life to cosmology through Titan-like environments. The cosmic microwave background cooled according to

$B$3

so that at redshift $B$4 the CMB temperature equaled Titan’s present-day surface temperature of about $B$5 K. The argument is that Titan-like bodies in the earliest metal-rich galaxies could have maintained this temperature for tens of Myr irrespective of their distance from a star, potentially allowing life to emerge merely $B$6 Myr after the Big Bang. Titan is central because it may host both a subsurface water ocean relevant to life-as-we-know-it and rivers, lakes, and seas of liquid methane and ethane relevant to life-as-we-do-not-know-it (Loeb, 2022).

A third scenario shifts the locus of origin from Earth to Mars. In that account, a cyanosulfidic origin-of-life chemistry requiring ultraviolet light, wet-dry cycling, and volcanism may have been more plausible on early Mars than on an early Earth that was likely an Ocean World with little exposed continental crust. The same paper argues that the timing of the great oxidation event on Earth, around $B$7 Ga, is inconsistent with final fixation of the genetic code in response to oxidative stress, whereas Mars may have become oxidizing earlier, especially around the Noachian-Hesperian boundary. This scenario combines cyanosulfidic photoredox chemistry with lithopanspermia from Mars to Earth (Carr, 2021).

A more general review of rocky planets organizes these questions around prebiotic chemical scenarios. It emphasizes that the liquid-water habitable zone is not identical to an abiogenesis zone. For cyanosulfidic chemistry, ultraviolet requirements may exclude some M-dwarf planets even if they are habitable in the liquid-water sense. The same review treats Titan’s methane lakes, Europa and Enceladus as ocean worlds, Mars as a record of ancient prebiotic conditions, and exoplanet populations as statistical tests of abiogenesis rate (Rimmer et al., 2021).

These origin narratives are not equivalent. One depends on genome-complexity extrapolation, one on early-universe thermal conditions around Titan-like worlds, one on Martian surface chemistry and transfer, and one on environment-specific prebiotic pathways on rocky planets. Their coexistence in the literature underscores that origin-of-life theory remains model-dependent and strongly conditioned by assumptions about chemistry, geology, and timescale (Sharov et al., 2013, Loeb, 2022, Carr, 2021, Rimmer et al., 2021).

4. Astrobiological detection and biosignatures

The search for life beyond Earth is presented not only as a search for biosignatures after life exists, but also as a search for prebiosignatures and planetary contexts favorable to abiogenesis. One framework proposes a workflow from laboratory chemistry, to in situ solar-system exploration, to exoplanet observations. In that hierarchy, lab work tests whether proposed pathways actually work, nearby worlds supply real planetary test beds, and exoplanet spectra provide the statistical context needed to constrain how often life arises (Rimmer et al., 2021).

Within that strategy, prebiosignatures are divided into primary and secondary classes. Hydrogen cyanide is the clearest example because it is both a reactant in many prebiotic pathways and a tracer of impacts, lightning chemistry in reduced atmospheres, and chemistry driven by stellar energetic particles. The same review identifies the coexistence of $B$8 or $B$9 with $q_1(x), q_2(x), \ldots, q_n(x)$0 as the gold standard atmospheric biosignature, while also listing multiple oxygen false-positive pathways, including runaway greenhouse scenarios, photodissociation on M-dwarf planets, and low-pressure atmospheres lacking cold traps (Rimmer et al., 2021).

A thermodynamic biosignature approach argues that astrobiologists should screen candidate planets not primarily for Earth-like morphology, but for fluids capable of supporting chemistry and compartment formation, evidence of thermodynamic disequilibrium, signs of synthesis, self-assembly, and accumulation, and especially increase in inhomogeneity over time. In that account, populated planets are dynamic inhomogeneous systems because stellar input drives disequilibrium externally while animate matter contributes to disequilibrium internally through accumulation and self-assembly (Popovic, 2018).

For Venus, the “Venus Life Equation” introduces a theory-and-evidence-based scaffold

$q_1(x), q_2(x), \ldots, q_n(x)$1

where $q_1(x), q_2(x), \ldots, q_n(x)$2 is Origination, $q_1(x), q_2(x), \ldots, q_n(x)$3 is Robustness, and $q_1(x), q_2(x), \ldots, q_n(x)$4 is Continuity. Origination is further decomposed into abiogenesis and panspermia, robustness into biomass and diversity, and continuity into the persistence of habitable conditions through time and space. The framework is explicitly informal rather than a rigorous statistical likelihood function, but it is intended to identify which poorly understood aspects of Venus can be constrained by future missions (Izenberg et al., 2020).

A recent chemically agnostic biosignature proposal focuses on amino-acid reactivity distributions rather than amino-acid presence alone. LUMOS, “Life Unveiled via Molecular Orbital Signatures,” analyzes abundance-weighted HOMO-LUMO gap values of amino acids in a sample. Abiotic samples display highly uniform distributions of amino-acid HOMO-LUMO gaps, whereas biotic samples show greater variance and preference toward lower HLG. The paper reports $q_1(x), q_2(x), \ldots, q_n(x)$5 accuracy in distinguishing biotic versus abiotic provenance, a ROC AUC of $q_1(x), q_2(x), \ldots, q_n(x)$6 for HLG weighted variance, and $q_1(x), q_2(x), \ldots, q_n(x)$7 accuracy with a coarse decision tree. It also notes that samples with fewer than about $q_1(x), q_2(x), \ldots, q_n(x)$8 amino acid species may be unreliable (Ramírez-Colón et al., 11 Feb 2026).

Mars remains a major target because missions can test for ancient habitability, organics, and oxidation history. The Mars-origin scenario highlights Jezero Crater as an open-basin lake system with deltas, sedimentary deposits, clay minerals, hydrated silica, possible carbonates, and stratigraphy that may record whether Mars became progressively more oxidizing through time. Titan is likewise treated as a natural laboratory because a positive detection there would bear on both non-aqueous habitability and early-universe habitability arguments (Carr, 2021, Loeb, 2022).

5. Artificial life, hybrid life, and AI-challenged boundaries

Artificial Life, commonly abbreviated as ALife, studies life as it could be rather than only life as it is on Earth. The field is explicitly interdisciplinary and seeks general properties of living systems without requiring terrestrial chemistry. Among the properties emphasized are homeostasis, self-reproduction, evolution, autonomy, robustness, self-organization, adaptation, self-repair, energy efficiency, materials recycling, local intelligence, and self-replication. A stronger criterion within this literature is autopoiesis: a system that self-produces and maintains itself (Gershenson et al., 2021).

The standard ALife tripartition is soft, hard, and wet. Soft ALife includes cellular automata, random Boolean networks, boids, artificial chemistries, swarm chemistry, digital evolution, and digital organisms. Hard ALife studies embodied machines and robotics. Wet ALife focuses on artificial cells, protocells, droplets, vesicles, micelles, and synthetic cells. Wet ALife is described as the most promising route to systems that people would plausibly call truly alive, but the same survey notes that an artificial cell that self-produces and maintains itself has not yet been demonstrated (Gershenson et al., 2021).

“Hybrid Life” extends this agenda to systems integrating biological, artificial, and cognitive components. Its three themes are theories of systems and agents, hybrid augmentation, and hybrid interactions among artificial and biological systems. Theoretical sources include cybernetics, Randall Beer’s dynamical-systems approach, autopoiesis, Integrated Information Theory, the Free Energy Principle, active inference, predictive information, and semantic information. The program treats life less as a fixed essence than as a pattern of self-sustaining organization, agency, autonomy, cognition, and interaction across biological and artificial boundaries (Baltieri et al., 2022).

Artificial intelligence further destabilizes traditional definitions. One review argues that advanced AI could exhibit autonomous learning, problem-solving, adaptation, self-improvement, self-replication, and possibly consciousness or self-awareness. It does not endorse AI as life outright, but it treats sufficiently advanced computational systems as a serious conceptual possibility for nascent life-like status. The same review separates life, intelligence, and self-awareness, emphasizing that human beings are unusual not merely for being alive, but for being self-aware that they are alive (Jiang et al., 2023).

This literature also records a methodological shift. Instead of asking only whether a system is biologically alive, recent work asks how systems maintain identity, how they act under precarious conditions, how they integrate memory and control, and how biological and artificial components can be coupled into new forms of organization (Gershenson et al., 2021, Baltieri et al., 2022).

6. Formal, mathematical, and engineering usages of “LIFE”

In formal systems, “Life” also names a family of cellular automata and related computational models. Conway’s Game of LIFE is the outer-totalistic binary cellular automaton with Moore neighborhood and rule $q_1(x), q_2(x), \ldots, q_n(x)$9. Mean-field analysis predicts an absorbing vacuum phase at $A$0, a stable active phase at $A$1, and an unstable saddle point at $A$2, whereas square-lattice simulations yield a much smaller stationary density $A$3. Reia and Kinouchi therefore interpret LIFE not as a system at the border of chaos, but as a quasicritical nucleation process on the border of extinction, with vacuum bubbles invading a metastable alive phase; for LIFE they report $A$4 (Reia et al., 2014).

Graph-theoretic generalization appears in “Life-like network automata” (LLNAs), where node updates depend on their own binary state and the fraction of live neighbors. In this formalism, the Game of Life becomes $A$5. The central conceptual move is a genotype/phenotype split: genotype is captured by the mean field curve and Derrida curve, while phenotype is captured by the state average $A$6 and defect average $A$7. The paper reports Pearson correlations of about $A$8 between $A$9 and $Q$0, and $Q$1 between $Q$2 and $Q$3. In a firing squad synchronization problem, this metric-based screening reduces rule space to $Q$4 candidates in $Q$5, with a best reported success rate of $Q$6 in the baseline setting and up to $Q$7 on small-world networks with $Q$8 and rewiring probability $Q$9 (Rollier et al., 26 Jun 2025).

A quantum adaptation assigns a qubit to each cell, so liveness becomes a probability density rather than a binary variable. In simulations on a $f:[a,b]\to[0,1], \qquad f(a)=0,\quad f(b)=1,$0 lattice with periodic boundary conditions, the system evolves toward a dynamic equilibrium with mean cell liveness

$f:[a,b]\to[0,1], \qquad f(a)=0,\quad f(b)=1,$1

and Gaussian standard deviation

$f:[a,b]\to[0,1], \qquad f(a)=0,\quad f(b)=1,$2

The model contains classical lifeforms, previously unseen quantum multicellular lifeforms, and $f:[a,b]\to[0,1], \qquad f(a)=0,\quad f(b)=1,$3 species of quantum lifeform. A qutub can be dormant or hibernating and, when destabilized, acts as a seed that can generate classical and quantum children, oscillators, a liveness probability density, or death, with outcomes highly sensitive to the initial state (Faux et al., 2019).

In engineering usage, LIFE can also denote a named AI framework rather than a theory of biological life. “LIFE — an energy efficient advanced continual learning agentic AI framework for frontier systems” expands to “Learning framework that is Incremental, Flexible, and Energy efficient.” It is proposed as an agent-centric continual learning framework for HPC data centers and autonomous network operations, organized around four components: an orchestrator, Agentic Context Engineering, the Agent Memory System for Networks, and Information Lattice Learning. The paper grounds the framework in a closed-loop example of detecting and mitigating latency spikes in critical microservices running on a Kubernetes-like cluster, and explicitly contrasts the design with a single monolithic model (Lee et al., 14 Apr 2026).

Taken together, these formal and engineering usages do not collapse into the biological concept of life. They show, instead, that “Life” functions as a productive metaphor and technical label for rule-based emergence, adaptive dynamics, distributed memory, and continual learning in computational settings (Reia et al., 2014, Rollier et al., 26 Jun 2025, Faux et al., 2019, Lee et al., 14 Apr 2026).