CeleryKills

I picked this exchange up from the Facebook feed of a friend who posted a concise summary of why evolutionary biology remains one of the strongest and most thoroughly supported scientific frameworks in modern science. The post pointed to the fossil record, genetics, DNA, and direct observation, then ended with a touch of exasperation:

“When comparing creation and evolution, the scientific consensus overwhelmingly concludes that evolution is the process that actually happened to bring about the diversity of life on Earth…”

Almost immediately, the discussion took a familiar turn. A commenter attempted to overturn the entire field with a probability argument and an appeal to artificial intelligence:

“According to Grok4 AI, The probability of a new functional gene around 1000 nucleotides forming via random mutations is about one in 10^600.”

The conclusion arrived right on schedule:

“The only way it can possibly happen is if an ultra intelligent being, who we call God, designs it.”

What caught my attention was not the number. It was the question that produced it.

I have seen versions of this argument for decades. Before Grok, before ChatGPT, before search engines, before social media, there was Fred Hoyle’s Boeing 747 analogy. Hoyle argued that life arising naturally was comparable to a tornado sweeping through a junkyard and assembling a fully functional airliner (Hoyle, The Intelligent Universe, 1983). It sounded devastating because it smuggled in a hidden assumption.

The assumption was that evolution works by random assembly toward a predetermined goal.

Evolution does not do that.

That distinction is everything.

My reply focused on precisely that missing model:

“If you ask for the odds of a specific 1,000-nucleotide gene appearing in a single leap through random mutations, of course you get an absurdly small number. But evolution is not a lottery where nature announces the winning sequence in advance and then waits for blind chance to hit it exactly.”

The phrase “blind chance” has done immense damage to public understanding of evolution. Random mutation exists. Natural selection exists. They are not the same thing.

A mountain stream is blind. It has no purpose, no intention, no destination in mind. Yet over time the stream carves channels through stone. Nobody imagines the water consciously designed the canyon. The process is unguided, but it is not arbitrary.

The same applies to evolution.

• Mutations occur.
• Some variants survive better.
• Some reproduce more effectively.
• Some persist long enough to become the foundation for future variation.

At no point is a future gene sequence specified in advance. There is no cosmic answer key hidden at the back of the book waiting to be matched exactly (Dawkins, The Blind Watchmaker, 1986; Mayr, What Evolution Is, 2001).

That is why the probability argument fails.

It calculates the odds of a process evolutionary biology does not propose.

The argument imagines a modern gene appearing all at once, fully formed, by pure chance.

You can also see the same tactic being taken in more recent works like Ted Beales (aka Vox Day) book Probability Zero (Beales, 2026, Probability Zero). Beale does not include: selection, drift, recombination, gene duplication, and population structure. He assumes a predetermined destination, does not allow for intermediate functional states, and has not been subjected to any kind of peer-review.

Evolution proposes cumulative selection, retention of successful variants, exaptation of existing structures, gene duplication, recombination, and countless incremental modifications across vast spans of time (Ohno, Evolution by Gene Duplication, 1970; Lynch, The Origins of Genome Architecture, 2007).

Functional Genes Are Not Isolated Targets

The probability argument also depends on a picture of genetics that has not survived contact with modern biology. The familiar calculation assumes there is one correct 1,000-base-pair sequence hiding somewhere within an ocean of nonfunctional possibilities. Miss a nucleotide and the entire exercise fails. That image resembles a combination lock with a single winning combination.

Real genes do not behave that way.

The genetic code itself contains redundancy. Multiple codons can encode the same amino acid, meaning that many different DNA sequences can produce identical protein sequences. Even at the protein level, organisms routinely tolerate substitutions, insertions, and deletions while retaining function. Entire families of proteins can differ at dozens or even hundreds of positions and still perform the same biological role (Alberts et al., Molecular Biology of the Cell, 2014; Wagner, Arrival of the Fittest, 2015).

What emerges is not a landscape dotted with isolated islands of functionality. It is a vast network of interconnected possibilities. Evolution is not searching for one winning lottery ticket. It is exploring broad regions of sequence space containing countless acceptable solutions. The probability argument quietly assumes a dartboard with a single microscopic bullseye. Biology keeps revealing a target that is much larger and far more forgiving.

The destination does not exist beforehand.

Only the next step matters.

Evolution Builds With Existing Materials

Much of the argument also imagines new genes appearing from scratch as though nature periodically sits down at an empty workbench and begins typing an entirely new instruction set.

Evolution rarely works that way.

One of its most powerful mechanisms is gene duplication. A gene is copied. One copy continues performing its original role while the duplicate is free to accumulate changes. Over time that duplicate may specialize, acquire a new function, become active in different tissues, or develop entirely new capabilities (Ohno, Evolution by Gene Duplication, 1970; Lynch, The Origins of Genome Architecture, 2007).

This process changes the problem dramatically. Instead of producing a complex genetic sequence from nothing, evolution begins with a sequence that already functions. Modification replaces invention.

What appears at first glance to be a miraculous leap often turns out to be a long series of small edits applied to existing biological text.

Nature Recycles More Than It Invents

Exaptation complicates the probability argument even further. Biological structures frequently acquire functions very different from those for which they originally evolved.

Feathers provide one of the best-known examples. Evidence suggests that feathers were present before powered flight and likely served purposes involving insulation, display, or other functions before becoming useful for flying (Brusatte, The Rise and Fall of the Dinosaurs, 2018). Similar patterns appear throughout biology. Existing proteins become part of new molecular systems. Existing structures find new uses.

I often think this is where engineering analogies begin to mislead people. Engineers typically start with a goal and work toward it. Evolution starts with whatever happens to be available and repurposes it.

Nature looks less like an architect following blueprints and more like an endlessly resourceful tinkerer rummaging through an already crowded workshop.

The Hidden Importance of Neutral Change

There is another feature almost completely absent from these probability calculations.

Many mutations are neither obviously beneficial nor harmful.

They spread through populations through genetic drift, creating reservoirs of variation that can later become useful under different environmental conditions. Genomes continuously accumulate these neutral changes. Future adaptations frequently build upon them (Kimura, The Neutral Theory of Molecular Evolution, 1983).

The common creationist model assumes every genetic change must immediately justify its existence by producing a complete modern function.

Evolution does not operate under such restrictions.

The process can explore vast areas of sequence space long before any particular configuration becomes advantageous.

Selection Is Not Waiting For Perfection

Perhaps the largest flaw in the entire calculation is its assumption that complex adaptations must emerge fully formed.

Imagine an enzyme whose effectiveness improves gradually through multiple mutations. If each change provides a modest improvement, natural selection can preserve every successful stage. No giant leap is required. No final target is known in advance. No organism must somehow stumble onto the finished sequence in a single event (Dawkins, The Blind Watchmaker, 1986; Mayr, What Evolution Is, 2001).

This transforms the mathematics.

The relevant probability is no longer the chance of obtaining the final version all at once.

The relevant probability becomes the likelihood of achieving a series of incremental improvements, each preserved because it works better than what came before.

That is a very different problem.

• A sequence that performs slightly better survives.
• A sequence that survives becomes the starting point for future change.

The process is goal free.

That phrase matters because so much of the creationist argument quietly depends on importing human concepts into nature.

• Purpose.
• Intent.
• Design.
• Goals.
• Direction.

These are deeply anthropocentric categories.

Humans like goals. Humans construct plans. Humans build things for reasons. We naturally project those experiences onto the larger universe and then ask why the universe appears designed.

But remove intention as a requirement and many supposedly impossible puzzles become ordinary processes.

Remove contingent existence as a requirement and the universe no longer owes us an external explanation simply because human minds prefer one.

Remove intelligibility as a mystical property and mathematics becomes what it has always been: a powerful human toolkit developed to describe recurring patterns we observe in nature rather than evidence that the cosmos was written in a language intended for us (Wigner, The Unreasonable Effectiveness of Mathematics, 1960).

Summary: The Wrong Model Produces The Wrong Answer

The recurring pattern in these arguments is surprisingly consistent.

1. A specific modern gene is chosen as the target.
2. All alternative functional sequences are ignored.
3. Gene duplication is removed.
4. Exaptation is removed.
5. Neutral evolution is removed.
6. Cumulative selection is removed.

The final sequence is treated as the only acceptable outcome.

The resulting probability is then presented as though it describes evolution.

It does not.

It describes a process evolution abandoned before Darwin was born.

Blind processes are not necessarily blind chance. Natural selection, genetic drift, exaptation, duplication, and cumulative retention fundamentally change the search landscape. Once those mechanisms are restored, the famous astronomical probabilities stop describing evolutionary biology and begin describing a caricature of evolutionary biology.

The calculation survives only because the model underneath it has already removed the very mechanisms responsible for biological innovation.

This is where the Grok citation becomes especially interesting.

AI systems answer the questions they are given.

They do not automatically identify every hidden assumption embedded in a prompt.

Ask Grok for the probability of a modern gene appearing randomly in one gigantic leap and it will attempt to answer that question.

Something like this:

“Assuming a functional gene requires a specific sequence of approximately 1,000 nucleotides, what is the probability of that exact sequence arising through random mutations alone, without natural selection, recombination, gene duplication, or any cumulative evolutionary processes?”

or

“Calculate the probability that a completely new 1,000-base-pair functional gene could form by random chance from non-coding DNA if every nucleotide must be correct and no intermediate sequences provide any selective advantage.”

It will not necessarily stop and say:

“Wait. Evolutionary theory does not propose this model.”

That omission becomes the entire argument.

The resulting probability is then presented as though it refutes evolution when all it actually refutes is a caricature of evolution.

I find that remarkably revealing.

The conversation begins with a biological process that has no foresight, no objective, no designer, and no target sequence. Then it gets rewritten into a process that desperately needs all four.

The conclusion follows automatically because the assumptions were loaded before the calculation began.

In the end, the most important lesson has very little to do with Grok.

A perfect answer to the wrong question is still the wrong answer.

References

• Alberts, Bruce et al. Molecular Biology of the Cell. 2014.
• Brusatte, Steve. The Rise and Fall of the Dinosaurs. 2018.
• Dawkins, Richard. The Blind Watchmaker. 1986.
• Hoyle, Fred. The Intelligent Universe. 1983.
• Kimura, Motoo. The Neutral Theory of Molecular Evolution. 1983.
• Lynch, Michael. The Origins of Genome Architecture. 2007.
• Mayr, Ernst. What Evolution Is. 2001.
• Ohno, Susumu. Evolution by Gene Duplication. 1970.
• Wagner, Andreas. Arrival of the Fittest. 2015.
• Wigner, Eugene. The Unreasonable Effectiveness of Mathematics in the Natural Sciences. 1960.