The Dead End of Methodological Naturalism

 

Functional Intelligence, the Hollowing Out of the "Scientific" Label, and the Self-Contradiction of Reason

One of the most frequently repeated dogmas of modern philosophy of science is that methodological naturalism—which explains the world solely through natural, material causes and excludes an unseen Creator a priori—is a neutral, objective rule of inquiry. In reality, however, this stance is not a scientific conclusion derived from facts, but a preconceived worldview bias maintained by an academic system operating along human interests, existential constraints, and institutional dogmas.

When we examine reality not through prejudice, but by the standards of physics, computer science, and pure logic, methodological naturalism runs into severe self-contradictions from two directions: from the outside (regarding the measurable level of intelligence in the living world) and from the inside (regarding the reliability of human reason itself).

Photo by sebastiaan stam on Unsplash

1. Functional Intelligence as a Standard: The Absurdity of Denying Design

The analogical argument (the argument from design) states that because DNA is an information system just like software—in fact, infinitely more complex—it logically requires a designer just as artificial intelligence (AI) does. Naturalistic science, however, rejects this analogy, claiming that living organisms reproduce and mutate, and therefore cannot be compared to machines created by human engineers.

Yet this counterargument is false, because it ignores the fact that comparing AI to a living organism is not based on external, formal resemblance (like a wooden horse versus a living horse), but on identical operational and information-processing principles. Both artificial intelligence and a living organism are functioning, information-processing entities.

When we compare the operational parameters of the two systems against the measurable facts of physics and computer science, the denial of design becomes utterly absurd:

Data Density and Storage Capacity (DNA vs. Silicon): Modern data centers fill massive server halls to operate AI. In contrast, biochemical research has demonstrated that the theoretical storage capacity of a single gram of DNA reaches 215 petabytes (215 million gigabytes)—meaning the entirety of humanity’s digital knowledge could fit into a single teaspoon of biological storage medium, in a 3D, self-repairing format.

Energy Efficiency and Cooling (20 Watts vs. Power-Plant Scale): Training a single state-of-the-art language model takes weeks or months, with tens of thousands of graphics processing units (GPUs) running continuously at peak capacity. The instantaneous power draw of such a data center typically ranges between 20 and 100 megawatts (MW), and tech giants are already building 1-gigawatt (1,000 MW) facilities—matching the capacity of a small nuclear power reactor. Over time, the energy required for hundreds of millions of daily queries exceeds even the energy cost of training, and nearly 100% of the consumed electricity is converted into heat: the evaporative cooling towers of a medium-sized data center consume millions of liters of freshwater daily (generating just 20 to 50 short message exchanges evaporates roughly half a liter of water). By contrast, the human brain—or that of any living mammal—operating billions of neurons and trillions of synapses in real time while controlling movement, circulation, the endocrine system, and conscious thought, runs on a mere 20 to 25 watts of power—the energy demand of a single dim light bulb—without water-cooling towers.

The Prerequisite of Learning (Hardware and Training): The human brain learns over decades, in small steps, through everyday life; however, the learning process has nothing to do with the development of the brain itself, because unless the brain is already a perfectly functioning system from the outset, there is nothing with which to learn. Likewise, one first needs a top-tier model before it can be trained. Yet while science claims that intelligent design is required for both the model and its training, it insists none is needed for the brain—claiming instead that it forms itself like a snowball rolling down a snowy slope, shaped by the environment like an avalanche until it turns into a snowman.

Self-Organizing Hardware Manufacturing (Nanoscale Engineering): AI software is incapable of building new servers and chips for itself; doing so requires external factories, mining, and human engineers. A living organism, however, starting from a single fertilized egg cell, constructs its own hardware: on the molecular assembly lines of ribosomes, it assembles thousands of proteins per minute based on the DNA code, while cells differentiate into bone, muscle, neural pathways, and a circulatory system—all without an external assembly plant.

Real-Time, Autonomous Error Correction: While the code of AI systems must be continuously updated and debugged by human programmers, inside a living cell, DNA polymerase enzymes and mismatch repair protein complexes inspect and correct the code during replication with an accuracy of 1 error in 10 billion, while irreparably damaged units are immediately decommissioned through programmed cell death (apoptosis).

ATP Synthase (The World’s Smallest and Most Efficient Rotary Motor): While the best electric motors designed by human engineers achieve an efficiency of around 85–95% (and internal combustion engines only 30–40%), the ATP synthase operating inside the mitochondria of living cells is a genuine rotary nanomotor complete with a shaft, a stator, and a rotor. It spins at 100 to 300 revolutions per second, produces exactly 3 ATP molecules per full rotation, and converts electrochemical energy from proton flow into mechanical rotation and then into chemical bonds with nearly 100% thermodynamic efficiency—a feat unattainable by human engineering. Moreover, it is capable of bidirectional operation (reversibility): when needed, it reverses its rotation and consumes ATP to pump protons back across the membrane, functioning just like an intelligent two-way hydroelectric power plant.

The Ribosome (The Molecular 3D Printer and Translator): The ribosome is an autonomous, nanoscale production line that translates digital code copied from DNA (mRNA) into physical reality (proteins). It links 15 to 20 amino acids per second into a precise polypeptide chain with an astonishing translation accuracy of roughly 1 error in 10,000 amino acids, continuously supervised by built-in quality-control mechanisms (such as the "double sieve" of tRNA synthetase enzymes). Because the ribosome itself is composed of dozens of specialized proteins and RNA strands that can only be manufactured by an already functioning ribosome based on the DNA code, hardware and software are inseparably interdependent here (a classic chicken-and-egg problem), biochemically ruling out any blind, step-by-step assembly.

The Logical Self-Contradiction: The metabolism and error correction of a living cell, along with the synchronized operation of the nervous and circulatory systems, exhibit the exact same goal-oriented, problem-solving intelligence as artificial intelligence—only on an exponentially higher level. If the lower level of intelligence (AI) can only be explained by conscious design, then excluding a designer and appealing to blind chance in the case of an exponentially higher level of intelligence (the living organism) is a profound logical absurdity—regardless of whether the system runs on silicon hardware or carbon-based biology.

2. Linguistic Relabeling and the Hollowing Out of the "Scientific" Label

When methodological naturalism is confronted with these facts, instead of offering a genuine explanation, it resorts to a linguistic escape hatch: it invents the concept of teleonomy (apparent purposefulness). According to this view, a living system behaves as if it were designed, while in reality it is merely the product of blind chemical affinity and natural selection. However, calling a system that corrects errors in real time, builds its own hardware through self-organization, and converts energy at nearly 100% efficiency merely "apparently" purposeful is not a scientific proof—it is a conceptual shield.

Even more fundamentally, the very label "scientific" has been hollowed out, and today it signifies ideological loyalty rather than unbiased truth-seeking:

The Original Purpose of Science and the Betrayal of Truth-Seeking: Genuine science cannot proceed from preconceived worldview dogmas, because if a system rests on an a priori presupposition, the entire system is inherently distorted and false. The true, pure essence of scientific inquiry is the unbiased investigation of reality, wherever the evidence may lead—not the defense of a pre-established materialistic worldview.

Redefining the Term (Circular Reasoning): In the 19th and 20th centuries, materialistic philosophy seized a monopoly over the word "science" and rewrote its definition. From that point on, "scientific" no longer meant what is demonstrable and logical, but solely what fits within a materialistic framework. This is classic circular reasoning: "Science finds no evidence for a Creator because science, by definition, excludes a Creator."

The Fallacy of Appeal to Authority: In public discourse today, the word "scientific" acts like an official stamp that immediately shuts down debate. When a mathematical impossibility (such as the spontaneous origin of life) is stamped with the label "scientific hypothesis," the majority of people accept it as an established fact, when in reality it is an unproven article of faith.

Rejecting Observable Reality: While everyone in computer science and engineering knows that complex code and functional machinery always originate from an intelligent source, applying that exact same principle to biology is branded "unscientific." Within this closed system, an extraordinarily improbable material explanation is deemed "scientific," whereas a logically and mathematically far more probable non-material (intelligent) explanation is dismissed as "outside science." Thus, the label no longer describes reality; it protects the system from unwelcome conclusions.

3. Internal Collapse: Alvin Plantinga’s Evolutionary Argument Against Naturalism

Methodological naturalism not only fails to explain the intelligence of the living cell, it also invalidates scientific reasoning itself. Philosopher Alvin Plantinga demonstrated this in his famous thought experiment (Evolutionary Argument Against Naturalism):

Survival Is Not Identical to Truth: According to naturalistic evolutionary theory, the human brain developed through natural selection. However, natural selection cares solely about survival-enhancing behavior (such as fleeing from a predator) and is completely indifferent to whether the beliefs inside the brain are true.

False Beliefs Can Also Keep Organisms Alive: A prehistoric human runs away from a tiger just as effectively if he accurately recognizes the danger as he does if he falsely believes the tiger is a giant kitty he must race against, or an evil spirit he must flee from. Because an infinite number of false beliefs can produce successful survival behavior, under blind evolution the probability that our brains reliably produce true beliefs is exceedingly low.

The Logical Own-Goal: If naturalism and blind evolution are true, we have no reason to trust that the human mind is capable of discerning reality. And if our minds are unreliable, we cannot trust any theory produced by them—including naturalism and evolutionary theory themselves.

Conclusion: A Two-Way Dead End

Methodological naturalism thus traps itself in a twofold self-contradiction:

Looking outward, it denies that reason and a Creator stand behind living information systems that represent the pinnacle of engineering, attempting to mask the lack of explanation with linguistic relabeling ("apparent purposefulness") while declaring mathematical impossibilities to be "scientific."

Looking inward, it views the human mind as the product of a blind, purposeless process, thereby stripping its own scientific theories of any claim to truth.

When a system classifies claims as "scientific" that contradict probability theory, information science, and common sense—solely because they appeal to material causes—the label no longer represents science, but a closed ideology. In such a situation, true intellectual courage and scientific integrity mean refusing to bow to the label and instead following the facts and pure logic.

Trust in scientific knowledge stands on solid ground only when we acknowledge the fact to which both biochemistry and logic point: the visible world and the reason operating within it were created by the invisible Creator, proven by the intelligence necessarily commensurate with the information density found in the universe. If the incomprehensible quantity and quality of information present in the universe does not prove a personal intelligence proportionate to it, then science blatantly denies and repudiates the very parameters of rational thought.

The greater the light, the greater the darkness required to deny it. And this is precisely what methodological naturalism reflects—shutting itself off from obvious truth like the senseless, stubborn horse and mule mentioned in Psalm 32:9, which refuse to draw near to reality no matter how they are led.

The Real Cost of Artificial Intelligence and Solutions for the Future: From Power-Plant Scale to Neuromorphic Chips

Training and continuously serving state-of-the-art large language models (LLMs) does not happen out of thin air in the cloud; it requires massive, industrial and power-plant-scale physical infrastructure. Behind the computing capacity built by developers lies staggering energy and water consumption, especially when compared to the functioning of the human brain—which is precisely why engineers are now trying to copy biology itself to reduce this burden.

1. The Power-Plant Scale of Hardware and Training

Continuous Peak Load: Training a single top-tier AI model takes weeks or months, with tens of thousands of high-performance graphics processing units (GPUs) running continuously at peak capacity.

Nuclear Power Plant Output: The instantaneous electrical power draw of such an AI data center typically ranges between 20 and 100 megawatts (MW), while tech giants are already building 1-gigawatt (1,000 MW) facilities—matching the capacity of a small nuclear power reactor.

The Cumulative Cost of Daily Use: Although answering a single query requires a fraction of the energy compared to training (consuming roughly ten times as much electricity as a traditional web search), hundreds of millions of daily user requests add up over time to exceed the energy demand of training itself.

2. Heat Generation and Invisible Water Consumption

100% Heat Conversion: Inside the silicon circuits of servers, the flow of electrons encounters resistance, converting nearly all consumed electrical energy into heat. If the chips were not cooled immediately, they would overheat and fail within seconds.

Evaporative Cooling Towers: Many data centers use evaporative cooling towers. Heat drawn from the chips is transferred via water, which is exposed to the open air where a portion of it evaporates, thereby cooling the remaining water in the system. This evaporated water escapes into the atmosphere as vapor, meaning it is genuinely depleted from the local region's freshwater supply.

Half a Liter of Water per Conversation: A medium-sized AI data center consumes millions of liters of freshwater daily. On statistical average, generating just 20 to 50 short message exchanges evaporates roughly half a liter of water on the server side (depending on the data center's geographical location and local weather).

3. The Sharp Contrast: Why Is the Human Brain So Much More Efficient?

Different Operating Principles: While AI chips perform brute-force mathematical calculations to generate every single word (running operations across billions or trillions of parameters), the human brain operates through the movement of ions and chemical messengers (neurotransmitters), generating negligible heat.

Targeted Activation: When a human answers a question, only the necessary regions of the brain activate selectively, whereas a language model mobilizes massive parameter sets for every single response.

20 Watts vs. Megawatts: The human brain—or that of any living mammal—operating billions of neurons and trillions of synapses in real time while controlling movement, circulation, the endocrine system, and conscious thought, runs on a mere 20 to 25 watts of power—the energy demand of a single dim light bulb—without water-cooling towers.

4. Where Is the Trade-Off?

Speed and Scalability: The human brain is extraordinarily energy-efficient, yet translating a long foreign-language text or summarizing thousands of pages of data would take it hours, and it can only focus on one conversation at a time. AI, on the other hand, completes massive data-processing tasks in seconds and serves millions of people simultaneously—and the price of this industrial-scale speed is power-plant-level electricity and water demand.

5. How Are Developers Trying to Reduce This Burden?

Tech companies and engineers have recognized that traditional cooling and computing methods are unsustainable, so they are pursuing solutions in three main directions:

Closed-Loop Liquid Cooling (Direct-to-Chip and Immersion Cooling): Instead of evaporative towers, modern data centers circulate coolant through closed pipe systems directly over the chips, or submerge entire servers in specialized, non-conductive dielectric fluid. Because the system is closed, no water evaporates, resulting in virtually zero water consumption after the initial fill—and the captured heat can even be reused for district heating in nearby buildings.

Software Optimization (Sparse Activation – Mixture of Experts): Developers now design models so that when answering a query, the entire network of hundreds of billions of parameters does not fire at once; instead, only smaller "expert" sub-networks relevant to the specific topic are activated. This mimics the human brain's targeted regional activation and slashes the energy required per response to a fraction.

Neuromorphic Chips (Copying the Brain's Architecture): In conventional computers, memory and processing are separated, and constantly shuttling data back and forth consumes immense amounts of power. Neuromorphic chips, by contrast, imitate human neurons and synapses: data processing and storage happen in the exact same place, and the chip consumes power only when an actual signal (spike) passes through it. These systems can be up to a thousand times more energy-efficient than today's GPUs.

Tech companies, engineers, and developers are all hard at work here, yet we are told to strip any background intelligence out of raw natural processes. Still, the latter took home the prize? Let every person blessed with plain common sense reflect on that!




Nincsenek megjegyzések:

Megjegyzés küldése

The scientific refutation of darwinian evolution