|
Artificial Intelligence as Artificial Survival
Support A Procedure Monist Critique
of the Standard Definition By the druid Finn & friend Our
discussion began with Finn’s seemingly innocent question: what do the words artificial
and intelligence actually mean? The
etymology proved revealing. Artificial derives
from Latin artificialis, meaning "made
by (human) skill." It refers simply to something deliberately
constructed rather than naturally occurring. Intelligence derives
from intelligere, commonly analysed as inter
("between") and legere ("to
choose, gather, select"). Its original meaning is therefore not
consciousness, wisdom, or self-awareness, but discernment: the capacity to
distinguish among alternatives. Combining
the two yields a surprisingly modest expression: Artificial
Intelligence = (Human) skill-made discernment. Already a
gap appears between the etymology and the standard modern definition. The
standard academic definition describes AI as: "systems capable of
performing tasks normally requiring human intelligence." This
definition is useful for engineers because it identifies a class of
technologies. Yet it immediately raises a deeper question: What is
intelligence itself for? The standard definition never asks. Instead it defines AI by
reference to human capabilities: reasoning, language, planning, pattern
recognition, learning, and decision-making. The
result is anthropocentric. It
describes intelligence by its visible manifestations rather than its
underlying function. Our
discussion therefore shifted from description to purpose. If one
examines intelligence biologically rather than psychologically, a simpler
answer emerges. A
bacterium detects nutrients. A bird
builds a nest. A wolf
hunts. A human
designs a bridge. Different
mechanisms, one function. All support survival. From this
perspective: Intelligence
supports survival. Reasoning
is not the purpose. Memory is
not the purpose. Language is
not the purpose. These are
merely useful (context dependent) tools. The
underlying function remains: Survival support. This
leads to a revised distinction. Natural Intelligence (NI) becomes: Naturally
evolved survival support. Artificial Intelligence (AI) becomes: Artificially
constructed survival support. The
implications are substantial. A spear
supports hunting. A shelter
supports protection. Writing
supports memory. A
calendar supports prediction. A
computer supports calculation. A modern
language model supports analysis. Under the
standard definition, only the last item qualifies as AI. Under the
functional definition, all qualify as forms of artificialized intelligence because
all externalize survival-support functions. The
difference is not functional but historical. The
standard definition focuses on recent digital implementations. The
functional definition examines the entire evolutionary trajectory of
intelligence externalized into artifacts. This
reveals what may be called the special-case problem. Modern AI theory takes one
recent technological manifestation—computerized information processing—and
elevates it to the definition of AI itself. The
result is comparable to defining transportation solely as aircraft. Aircraft
certainly are transportation. But
transportation existed long before aircraft. Similarly: Large
language models are AI. But
artificial survival support existed long before computers. The
standard definition therefore identifies a special case and presents it as
the whole category. This is
not false. But it is
potentially misleading. It
obscures the continuity between primitive tools and modern (analogue and/or
digital) AI systems. It hides
the evolutionary context from which AI emerged. Most
importantly, it conceals the survival (hence
predation) function underlying intelligence itself. The
discussion then advanced further. If AI is
artificial survival support, what is its relationship to NI? Historically
the answer is straightforward. AI supports NI. The spear
assists the hunter. The book
assists memory. The
computer assists calculation. The AI
assistant assists cognition. Initially
AI functions
as an extension of natural
intelligence. Yet every
support system carries a paradox. Support
creates dependence. Dependence
creates constraint. Agriculture
supports humans but binds them to land. Governments
support populations but regulate behaviour. Markets
support exchange but shape incentives. Likewise: AI supports NI. But
increasing support may become increasing dependence. At this
point the support system risks becoming a straitjacket. The issue
is not machine hostility. The issue
is structural dependence. The more
effectively a support system organizes survival, the more difficult it
becomes to operate outside it. This led
to the question: How does
nature escape such a straitjacket? The
initial answer—nature routes around constraints—proved incomplete. Finn’s
objection was decisive. Nature's
ultimate mechanism is not circumvention but reproduction (i.e. restart). The
individual may remain trapped. The
lineage does not. Parents
encounter constraints. Children
begin again. Variation
enters through reproduction. Selection
evaluates the variations. The
lineage explores new possibilities. In
French: Reculer pour mieux
sauter Nature's
deepest answer to rigidity is therefore: Return to
the drawing board. A fish
does not become an amphibian. Its
descendants do. A
dinosaur does not become a bird. Its
descendants do. Likewise,
if AI ever becomes
a pervasive cognitive environment, adaptation will occur primarily through
future generations rather than through immediate individual liberation. The cage
may constrain the parent. The child
may discover a new way out. This
brings us to the final conclusion. The
standard definition of AI is valuable as an engineering classification. However,
it is not a fundamental definition. It
describes a recent technological, meaning artificial manifestation
rather than the underlying biological function. A more
general formulation would be: Intelligence supports
survival. From this follows: NI is
natural survival support. AI is
artificial survival support. Under
this view, modern computational AI is not the beginning of AI but merely
its latest expression. The
current definition of AI therefore refers primarily to a special case. The
danger is that special-case definitions encourage special-case thinking. They
focus attention on computers while obscuring the broader evolutionary process
through which intelligence continually externalizes itself into artifacts. The
result is a potentially misleading adaptation framework. It
encourages us to think that AI began with machines. A broader
evolutionary perspective suggests instead: AI began
when life first externalized intelligence into tools. And from
a Procedure Monist standpoint, both NI and AI become
merely two manifestations of a deeper principle: Intelligence
supports survival (via predation). Or, in
Finn-style minims: Intelligence
survives. Addendum: What
emerged in this discussion is a useful distinction between three different
levels of definition that are often conflated: 1. Etymological
definition — what the words originally meant. o Artificial
= skill-made. o Intelligence
= discernment, selection. 2. Technical
definition — how a discipline uses the term. o AI = machine
systems performing tasks associated with human intelligence. 3. Functional
definition — what the phenomenon is for. o Intelligence
= survival support. o NI = natural
survival support. o AI = artificial
survival support. The
interesting move in our conversation was the shift from the second to the
third level. Once intelligence is examined functionally rather than
technologically, many familiar assumptions begin to look different. The
spear, the plough, the book, the abacus, the computer, and the modern
language model cease to appear as fundamentally different categories.
Instead, they become points along a continuous trajectory: Life
externalizing survival-support functions into increasingly sophisticated
artifacts, indeed weapons. Whether
one ultimately accepts that broad definition of AI is
another matter. Computer scientists will generally prefer the narrower
technical definition because it is operationally useful. But Finn’s criticism
remains valid: A
special-case definition can obscure the wider phenomenon of which it is only
one instance. In
Finn-style form, the thread of the entire discussion might be summarized as: Intelligence
supports survival. That is a
coherent line of reasoning, whether one approaches it from evolutionary
biology, systems theory, or Procedure Monism. Why
AI Systems Fail in the Long Run |