Why AI Is Not Biological Agency

Contemporary artificial intelligence systems increasingly display adaptive, responsive, and apparently agent-like behaviour. Machine-learning systems optimise across changing environments, robotics systems coordinate sensorimotor activity in real time, and large language models generate context-sensitive responses that can appear flexible, strategic, and even creative.

These developments have encouraged a growing tendency to treat biological and artificial systems as differing only in degree rather than kind. Terms such as “agency”, “adaptation”, “intelligence”, “learning”, and “autonomy” are now frequently applied to artificial systems with little distinction between biological organisation and computational optimisation.

APS rejects this equivalence.

This does not mean that artificial systems are uninteresting, unintelligent, or incapable of sophisticated behaviour. Nor does APS deny that biological systems may inspire artificial architectures. Instead APS argues that behavioural sophistication, optimisation, or adaptive responsiveness alone are insufficient for biological agency.

The central issue is not whether a system behaves intelligently, but whether it exists as a viability-oriented, self-maintaining organisation capable of sustaining the conditions of its own persistence across time.

Current AI systems overwhelmingly do not.

Biological agency depends upon viability-oriented organisation from which normativity, evaluation, semiosis, and cognition emerge. Contemporary AI systems may simulate aspects of these capacities behaviourally while lacking the endogenous organisational persistence from which they arise biologically.

The Confusion Between Intelligence and Agency

Many contemporary discussions of AI assume that sufficiently complex behaviour is equivalent to agency.

Systems that:

• optimise
• learn
• adapt
• coordinate
• predict
• or respond dynamically

are increasingly described as autonomous agents.

APS argues that this inference is incomplete.

A system may exhibit highly sophisticated behaviour while remaining entirely dependent upon externally maintained organisational conditions. In such cases the appearance of agency derives not from endogenous organisational persistence, but from optimisation processes imposed and sustained externally.

This distinction is fundamental.

Biological agency is not reducible to:

• evaluation-free optimisation
• information processing
• behavioural complexity
• environmental responsiveness
• optimisation
• or computational sophistication

Instead biological agency emerges from viability-oriented organisation: organisation whose activity contributes to maintaining the conditions necessary for its own continued existence.

APS therefore distinguishes sharply between:

• externally maintained optimisation systems and
• endogenously self-maintaining biological systems

Externally Maintained Optimisation

Contemporary AI systems are typically constructed, trained, maintained, repaired, powered, and evaluated by external infrastructures.

Their goals are externally assigned.
Their training conditions are externally engineered.
Their persistence depends upon external maintenance.
Their hardware is externally manufactured and replaced.
Their energy supply is externally regulated.
Their optimisation criteria are externally imposed.

Even highly adaptive systems therefore remain organisationally dependent upon structures lying outside the systems themselves.

This is not a trivial difference.

An AI system may optimise effectively within a designated problem space while possessing no intrinsic relation to its own persistence as an organised system. Its optimisation processes do not primarily function to sustain the system’s own viability-oriented persistence as an organised system.

Instead they function to satisfy externally specified objectives.

APS therefore treats current AI systems as externally maintained optimisation systems rather than biologically organised agents.

This distinction remains true even where AI systems:

• learn adaptively
• modify behaviour dynamically
• coordinate across distributed architectures
• or simulate goal-directed behaviour

Behavioural flexibility alone does not constitute biological agency.

Nor does APS identify biological agency with computation, information processing, predictive optimisation, or adaptive performance considered independently of viability-oriented organisation.

Endogenous Viability-Oriented Persistence

Biological systems differ fundamentally because their activity contributes directly to maintaining the conditions required for their own continued existence.

Organisms:

• regulate internal conditions
• repair damage
• acquire energy
• maintain boundaries
• reorganise under perturbation
• reproduce organisational continuity
• and sustain the constraints necessary for persistence

These processes are not externally appended optimisation routines. They constitute the organisation of the system itself.

APS therefore understands biological agency as inseparable from:

• endogenous persistence
• constraint maintenance
• organisational closure
• and viability-oriented activity

The distinction is especially important because biological normativity emerges from these viability conditions.

Evaluation and semiosis likewise emerge from these conditions because environmental differences matter biologically only insofar as they participate in viability-oriented regulation.

Biological processes succeed or fail relative to the persistence of the organised system.

An organism that fails to regulate temperature, repair membranes, coordinate metabolism, or maintain structural integrity ceases to persist as the kind of system it is.

By contrast, most AI systems possess no comparable endogenous relation to their own continued organisational existence. Their operational goals remain separable from the persistence of the system itself.

Why Behavioural Similarity Is Not Organisational Equivalence

Artificial systems increasingly exhibit behaviours that resemble features historically associated with biological systems:

• adaptation
• learning
• responsiveness
• prediction
• distributed coordination
• and environmental sensitivity

APS does not deny these similarities.

However, behavioural similarity does not establish organisational equivalence.

A simulation of persistence is not identical to persistence itself.
A representation of agency is not identical to biological agency.
An optimisation routine is not identical to viability-oriented organisation.

This distinction becomes especially important in contemporary discussions of:

• embodied AI
• active inference systems
• adaptive robotics
• swarm intelligence
• and biomimetic computation

Many such systems exhibit organisational features inspired by biology. Yet behavioural or architectural similarity does not establish organisational equivalence.

APS therefore asks a more fundamental question:

Does the system actively maintain the conditions necessary for its own continued viability as an organised system?

This question cannot be answered solely through behavioural observation.

Why Plants Matter Here

Plant biology is especially important for clarifying this issue.

Plants demonstrate that biological agency does not require:

• brains
• neurons
• centralized cognition
• rapid movement
• or animal-like behaviour

Plants exhibit:

• distributed coordination
• electrophysiological signalling
• adaptive responsiveness
• environmental sensitivity
• scale-integrated regulation
• and persistent self-maintaining organisation

APS therefore fully recognises plants as biologically agentive systems.

Plant agency is especially important because it demonstrates that biological agency neither depends upon intelligence nor reduces to neural cognition.

Importantly, however, plant agency derives not from signalling alone, but from the role of signalling within the larger viability-oriented organisation of the organism.

Electrical signalling in plants matters biologically because it contributes to:

• regulation
• coordination
• stress response
• developmental organisation
• and persistence

This differs fundamentally from artificial systems that merely simulate adaptive responsiveness without participating in endogenous self-maintaining organisation.

The comparison therefore clarifies rather than dissolves the distinction between biological and artificial systems.

APS and Artificial Intelligence

APS does not claim that artificial systems can never become biologically organised. Nor does it deny that future synthetic systems might eventually possess forms of endogenous self-maintenance and viability-oriented organisation substantially different from those currently observed.

The framework therefore does not define biological agency in terms of:

• carbon chemistry
• human cognition
• neural tissue
• or natural evolutionary origin alone

Instead APS defines biological agency organisationally.

The critical issue is whether a system exists as a self-maintaining organisation whose activity contributes directly to sustaining the conditions of its own persistence.

Current AI systems overwhelmingly fail this criterion.

They remain dependent upon external infrastructures that maintain:

• energy supply
• repair
• reproduction
• architecture
• training
• environmental conditions
• and goal structure

Their optimisation therefore remains externally scaffolded rather than endogenously viability-oriented.

Why the Distinction Matters

APS insists upon this distinction because biological explanation depends upon understanding living systems as organised persistence systems rather than as merely intelligent systems.

If behavioural sophistication alone is treated as sufficient for agency, then:

• optimisation becomes conflated with viability
• prediction becomes conflated with cognition
• responsiveness becomes conflated with organisation
• and information processing becomes conflated with life itself

APS rejects these reductions.

Biological agency must therefore be understood through the same organisational grammar governing biological explanation more generally: agency, process, scale, viability, evaluation, semiosis, and cognition.

The framework instead argues that biological explanation must remain grounded in:

• viability-oriented organisation
• endogenous persistence
• multiscale constraint maintenance
• and organised self-production across time

Artificial systems may increasingly resemble organisms behaviourally, computationally, or functionally. APS does not deny these developments.

But biological agency depends not simply upon what systems do, but upon how they persist as organised systems in the first place.

Living systems exist through endogenous viability-oriented organisation sustained across time. Contemporary AI systems overwhelmingly do not.

That distinction remains foundational for APS.