Something important may be hidden in the original reasoning, as with Brian Keating’s “natural nuclear reactor” case against intelligent design.

A few months ago, I came across a short video clip of a discussion between physicist and podcaster Brian Keating and philosopher of science Stephen C. Meyer. Keating presented Meyer with a rather intriguing challenge: If complex, functional systems require intelligent design, how do we explain the existence of a naturally occurring nuclear reactor?

Yes, you read that correctly. A natural nuclear reactor!

I found the argument fascinating and, admittedly, quite challenging. It sounded like a powerful objection to the idea that sophisticated functional systems require an intelligent origin.

But then I watched their full discussion, in which Meyer explained how probability allows us to infer the best explanation among competing possibilities. In the case of Oklo’s natural nuclear reactor, for example, natural processes provide the best explanation, whereas in other cases, intelligent design may be more probable.

I found this reasoning compelling. But something still bothered me: Probability, however small, always leaves room for chance. Could there be another way to approach the question, without having to rely on probabilities alone?

Then I remembered a lesson I had learned some years earlier: The devil is often hiding in the argument’s premises.

The art of turning an argument around

We have all experienced moments when someone presents an argument that sounds so convincing that we struggle to find a response. Perhaps the reasoning seems flawless, or perhaps the conclusion is so surprising that it leaves us questioning something we previously took for granted. But a convincing argument is not necessarily a sound argument.

Every argument rests on premises: assumptions or propositions that we accept, explicitly or implicitly, in order to arrive at a conclusion. And while we often focus on evaluating whether the conclusion follows, we sometimes neglect to examine the foundations supporting it. One technique I have found particularly useful is remarkably simple: Turn the argument around and ask the question in reverse.

By doing so, we can often uncover assumptions that were hidden in the original formulation. The purpose is not to win an argument or embarrass the person presenting it. Quite the opposite. The purpose is to move closer to the truth. Sometimes reversing an argument exposes a weakness in the argument’s reasoning. Other times, it reveals a weakness in our own understanding. Either way, we learn something new.

Let me illustrate this with two examples before returning to the curious case of the natural nuclear reactor.

Is science the only path to truth?

I first appreciated the power of this technique when listening to Oxford mathematician John Lennox. In one of his discussions, Lennox addressed a claim that is increasingly common in debates about science, philosophy, and religion: “Science is the only way to discover truth.”

At first glance, this sounds quite reasonable. After all, science has given us an extraordinary understanding of the physical world. From the structure of atoms to the evolution of galaxies, its achievements are undeniable.

But Lennox turns the question around. He asks: Can the statement “Science is the only way to discover truth” itself be scientifically established? What experiment could demonstrate it? What measurement would confirm that no other method of reasoning can lead us to truth?

The problem becomes immediately apparent. The statement is not itself a scientific discovery. It is a philosophical claim about the limits of knowledge. In other words, the claim appears to exclude itself from the very standard it demands of every other claim.

Notice what happened. We did not need to question the accomplishments of science, nor did we need to introduce an alternative theory of knowledge. We simply applied the proposed standard to the statement itself. And suddenly, an argument that initially sounded obvious became much less convincing.

Does absolute truth exist?

I have encountered a similar situation in discussions about morality. Occasionally, these conversations lead to a strongly materialistic interpretation of human existence: We are merely biological machines, responding to inputs according to physical laws. Our decisions are predetermined, moral values are human inventions, and ultimately nothing objectively matters. Such discussions sometimes rest on an even broader premise: “There is no absolute truth.”

But what happens if we turn that statement around? Is the claim that there is no absolute truth itself absolutely true? If it is, then at least one absolute truth exists, contradicting the original claim. If it is not, then we have no reason to accept the original claim as a universally true statement.

Of course, this observation does not by itself establish the existence of objective morality. That is a separate and much larger philosophical question. But it does reveal a problem with the premise often used to dismiss it.

Once again, the interesting part is not that we have defeated an opponent. It is that a simple reversal has exposed something important that was hidden in the original reasoning.

Now, let’s apply the same technique to the natural nuclear reactor at Oklo.

The nuclear reactor that nature built

Approximately two billion years ago, in what is now Gabon, Africa, a remarkable phenomenon occurred. A deposit of uranium, under the right geological conditions, sustained nuclear fission reactions. Groundwater acted as a neutron moderator, allowing a chain reaction to occur. As the water heated and evaporated, the reaction would slow or stop, only to resume when the conditions became favorable again. This was the Oklo natural nuclear reactor.

Long before humans understood nuclear physics, nature had already produced something capable of sustaining nuclear fission. In his discussion with Stephen Meyer, Brian Keating used this example to challenge the inference from sophisticated functionality to intelligent design. After all, if nature can produce something as remarkable as a nuclear reactor, why couldn’t it produce other complicated functional systems?

Meyer’s response centered on probability and inference to the best explanation. He argued that the relevant question is whether the natural processes available have sufficient probabilistic resources to produce the system in question. Some functional arrangements may be explained naturally, while others, he argued, are too improbable.

I find this response interesting, and probability certainly has a role in evaluating possible explanations. But I would like to approach the question differently. Instead of immediately calculating how unlikely a functional arrangement might be, why don’t we first turn Keating’s argument around?

How do we know that the Oklo reactor was formed naturally in the first place?

This may initially sound like a naive question, but let’s examine it for a moment. We have never observed the original formation of Oklo. Yet we confidently conclude that it was a natural phenomenon, rather than a primitive nuclear reactor built by some ancient or extraterrestrial civilization. But how can we be so confident? If nature can produce functional systems that resemble our own engineering, what allows us to distinguish the two?

Consider Stonehenge, for example

The very confidence with which we classify Oklo as natural suggests that we already rely on some criteria for distinguishing naturally occurring systems from engineered ones.

To make the contrast even starker, consider another ancient structure, Stonehenge, in England. We are equally confident that Stonehenge was engineered, even though we did not observe its construction and still debate its precise purpose. What makes us so confident in these two very different conclusions?

Both involve arrangements of naturally occurring materials. Both depend on specific physical conditions. And both exhibit forms of functionality. Yet we confidently recognize one as natural and the other as engineered, without having to explicitly calculate the probability of either arrangement.

If we can distinguish Oklo from Stonehenge with such confidence (without ever calculating the probability of either arrangement occurring naturally!), then there must be observable characteristics that justify treating one as naturally occurring and the other as engineered.

To further emphasize this distinction, let us imagine that tomorrow NASA announces an extraordinary discovery. A rover exploring Mars has encountered a structure nearly identical to Stonehenge. Large stones stand upright in a carefully arranged circle. Horizontal stones rest across them, forming interconnected structures. The stones have complementary shapes, with protrusions fitting into corresponding cavities.

Until this discovery, we had no evidence that intelligent life had ever existed on Mars. What would we conclude? I suspect that most scientists, and indeed most people, would immediately conclude that Mars once hosted an intelligent civilization capable of engineering such a structure.

We would not know who built it. We would not know why they built it. We might not even understand its ultimate purpose. Yet the organization of the stones alone would lead us to suspect engineering, even on a planet where we had no prior evidence of intelligent life. Why?

Organization, not mere complexity

To understand what makes this distinction so apparent, I decided to start with a more direct comparison: What distinguishes the natural nuclear reactor at Oklo from a modern nuclear reactor engineered by humans?

Consider a modern nuclear reactor. Its components are not simply placed together so that nuclear fission can occur. They are organized into a system designed to perform specific functions. The reactor accepts inputs, such as fuel, coolant, and control commands, and produces outputs, such as heat and information about its operating conditions.

But what makes this organization fundamentally different from the natural arrangement of uranium and groundwater at Oklo? I believe two characteristics are particularly revealing: the ability to distinguish successful operation from failure, and the ability to recognize and respond to specific inputs according to defined operational rules.

Functionality requires success and failure criteria

Let us start with what I think is perhaps the most important distinction between Oklo and an engineered nuclear reactor: What would it mean for either system to fail?

For a modern nuclear reactor, the answer is straightforward. Its operation must satisfy specific requirements, including acceptable temperature, pressure, and radiation levels. If its cooling system malfunctions, its temperature exceeds safety limits, or it cannot sustain its intended operation, we can identify a failure.

But what would it mean for Oklo to fail? If its nuclear chain reaction stopped, would that be a failure? If the surrounding groundwater disappeared, would that constitute a malfunction? What if the uranium deposit had never sustained a chain reaction in the first place?

There is no obvious answer. None of these outcomes can inherently be described as a success or failure. They are simply different physical states resulting from the interactions of uranium, groundwater, and the surrounding geological environment. Oklo had no operational specification defining what it was supposed to accomplish or which outcomes would constitute failures.

Now consider how this distinction extends beyond nuclear reactors. A computer fails when it cannot execute instructions correctly. A living cell fails to maintain its viability when essential biological processes break down, eventually leading to its death. In each case, we can identify what the system must accomplish and the conditions under which it fails to do so.

This reveals an important characteristic of functional systems: Their organization supports identifiable functions under a specification that distinguishes successful operation from failure.

Functionality requires defined inputs

The second characteristic of a functional system is its ability to recognize and respond to specific inputs according to defined operational rules.

Consider again our modern nuclear reactor. Its control system is organized to recognize particular commands and signals, such as instructions to adjust the control rods or measurements from temperature sensors. These inputs are interpreted by the system and trigger specific operations. But not every external signal or physical influence is a valid input. The system is configured to recognize particular inputs while ignoring others or treating them as disturbances.

Oklo, by contrast, responded to changes in its surrounding physical conditions. Groundwater, temperature, and uranium concentration influenced whether nuclear fission could be sustained. However, these interactions did not involve interpreting inputs according to an established set of operational rules. They were simply the result of physical interactions among the materials involved.

The distinction becomes even clearer when we consider computers. Every transistor in a computer follows the laws of physics. Yet the organization of these transistors determines which instructions the computer can recognize and execute. Two processors built from similar physical components can support entirely different instruction sets, such as x86 and ARM architectures. A sequence of bits that represents a valid instruction for one architecture may be meaningless to another. The difference is not in the physical properties of the bits, but in the organization of the system interpreting them.

Similarly, a living cell contains specialized molecular machinery capable of recognizing and processing particular inputs. Ribosomes, for example, interpret sequences of messenger RNA to synthesize proteins according to the genetic code. A sequence that the cellular machinery cannot recognize or process cannot serve as a functional instruction, regardless of the information it might contain.

The laws of physics determine what the components can do. The organization of these components determines which inputs the system can interpret and what operations it can perform.

Testing the distinction: A cave and a house

These functional distinctions are not unique to nuclear reactors. We can observe similar differences in other naturally occurring and engineered structures. Consider, for example, a natural cave and a house.

Both can provide shelter from rain, wind, and extreme weather. Yet if a cave has inadequate structural support, insufficient ventilation or natural light, or allows rainwater to enter, we would hardly call it a failure. A house with the same problems, however, would be considered poorly designed or even uninhabitable, because it fails to satisfy its functional requirements.

There is another interesting difference. Consider the dimensions of doorways, the placement of door handles, or the shapes of beds and chairs. These features are not arbitrary. They reveal that a house is organized to accommodate particular users and their physical characteristics. A doorway designed for an average adult human would hardly be suitable for an elephant! A cave, by contrast, has no corresponding features specifically configured for its occupants. Its suitability as shelter depends on whether an animal happens to fit within its natural dimensions, rather than the cave’s components being organized to accommodate that animal.

Now, back to Stonehenge

This brings us back to Stonehenge. Its individual stones are not merely arranged in a circle. They are shaped, positioned, and connected in ways that depend on one another. The builders employed mortise-and-tenon joints, in which protrusions on upright stones fit into corresponding holes in horizontal lintels. The lintels themselves were connected using tongue-and-groove joints.

These connections establish mutually dependent functional constraints. The dimensions and positions of the stones must be compatible for the structure to hold together. Change those dimensions arbitrarily, and the joints may no longer fit or the structure may lose its stability. Just as with a house, we can identify what the components must accomplish and the conditions under which their functionality would fail, even without knowing the monument’s ultimate purpose.

Where does meaning come from?

What started as a simple attempt to reverse a convincing argument and examine its premises has led us to a deeper understanding of what constitutes a functional system and the characteristics that distinguish it from naturally occurring phenomena.

What all these examples reveal is that functionality is not simply a matter of complexity, improbability, or even the ability to produce a useful outcome. It depends on how a system is organized to recognize and interpret particular inputs, perform specific operations in response, and satisfy identifiable constraints of success and failure. Whether we examine a nuclear reactor, a computer, a house, Stonehenge, or even a biological cell, we recognize functional relationships between components that are not explained merely by describing their individual physical properties.

But there is another important consequence of this observation. The same input can have entirely different functional meanings depending on the system receiving it. As we saw with computer architectures, a sequence of bits may represent a valid instruction for one processor but be meaningless to another. Similarly, DNA sequences acquire their biological functionality through the cellular machinery capable of interpreting them. In both cases, the information and the system interpreting it are inseparable from the functions they enable.

This raises a deeper question that I have been exploring in my research: Where does meaning come from?We often think of information as a sequence of symbols, but symbols alone have no functional meaning. It is the relationship between information and the system interpreting it that gives rise to functionality and meaning. As I explore in my article “One Question to Rule Them All: The Origin of Functional Information,” this question has implications far beyond nuclear reactors and ancient monuments. It touches on some of the most fundamental mysteries in science and philosophy, from the origin of life to the possibility of artificial general intelligence.

 Perhaps the deeper mystery is not simply where information comes from, but how systems capable of giving it meaning originated in the first place.