For over half a century, Antony Flew was a leading atheistic philosopher and was arguably the intellectual standard-bearer for modern atheism. His 1950 paper, “Theology and Falsification,” became one of the most cited philosophical works of the era, setting the terms of debate for generations of academics. When Flew announced in 2004 that he had come to accept the existence of a God, the reaction was understandably significant. Flew remained a rationalist, insistent on his lifelong Socratic principle of “following the evidence wherever it leads.” In his book, There is a God, Flew details the scientific evidence that he concluded had made materialism untenable and contributed to his acceptance of deism. This evidence fell into two main categories: the fine-tuning of the universe’s physical laws (cosmology) and the mystery of life’s origin (biology). While the physics provided a persuasive argument, it was the biological problem that Flew found insurmountable. The mathematical impossibility of DNA, with its specified, code-like information, arising from undirected, random processes, became the final, decisive obstacle to his atheism.

The “monkeys with typewriters” analogy, originating from concepts in statistical mechanics, served for decades as a popular and persuasive rhetorical tool. It was and still is an analogy frequently invoked to counter arguments from improbability, suggesting that any event with a non-zero probability must eventually occur given infinite time and resources. In this thought experiment, a monkey randomly hitting keys would eventually type the complete works of Shakespeare. Flew pointed to the issue of this analogy by citing a real-world experiment conducted in 2003. Funded by the British National Council of Arts, academics from Plymouth University placed a computer in the enclosure of six monkeys for a month. Flew noted the results: the monkeys produced fifty pages of text, consisting almost entirely of the letter ‘S’. Critically, they failed to produce a single word, not even the one-letter words ‘a’ or ‘i’ set apart by spaces. Furthermore, they quickly proceeded to urinate and defecate on the keyboard before one monkey attacked it with a stone, destroying it. This aptly demonstrates the issue that not only do you need to produce information, but you need to do so through brute forces that ensure the mechanism’s destruction. This practical failure merely highlighted the flaw in the abstract argument.

The analogy’s entire explanatory power relies on the endless availability of time and resources. The real universe, however, is finite. It has a finite age (estimated by proponents of this view to be 13.7 billion years) and a finite number of particles (estimated to be 10⁸⁰). These finite resources establish a concrete probabilistic “budget”: a maximum possible number of finite events, or “trials,” that could have ever occurred. More importantly, the timeframe suggested for abiogenesis is not the age of the universe. This same model suggests that life appeared on Earth relatively quickly, within a window of perhaps 500 million years after the planet cooled enough to support liquid water. This significantly shrinks the available time and resources for random chemical events. The monkey analogy, Flew concluded, was not a serious scientific argument but a convenient dismissal. It performs an intellectual sleight-of-hand, replacing a real-world finite impossibility with an abstract infinite inevitability.

Flew noted that the monkey analogy was misleading in a more fundamental way. The real problem is not that the monkeys fail to type something; they will produce endless strings of nonsense. This random gibberish has a kind of “complexity” but lacks specified complexity. A Shakespearean sonnet, by contrast, possesses specified complexity. It is a precise sequence of characters, arranged according to the rules of English grammar and syntax, to convey a specific meaning. Flew, referencing calculations by physicist Gerald Schroeder, highlights the statistical absurdity of achieving this by chance.

• Schroeder used Shakespeare’s Sonnet 18 (“Shall I compare thee to a summer’s day?”), which contains 488 letters. The probability of randomly typing this specific sequence is one in 26⁴⁸⁸, which equates to 10⁶⁹⁰.

This number represents the “cost” of the sonnet. To put this in context, one can calculate the universe’s total probabilistic resources or “budget.”

• There are an estimated 10⁸⁰ atoms in the universe.

• The fastest possible interaction is governed by Planck time (approximately 10⁴³ seconds), the smallest unit of time at which the laws of physics are thought to apply. This physical limit is crucial: it sets a hard ‘speed limit’ on reality, establishing a maximum number of ‘events’ or ‘state changes’ that can occur per second (roughly 10⁴³).

• The proposed age of the universe in seconds is 10¹⁷.

• Thus, the absolute maximum number of events that could have ever happened in the cosmos is 10⁸⁰ x 10⁴³ x 10¹⁷ which equals 10¹⁴⁰.

This number represents the universe’s total probabilistic “budget”. The comparison is confronting for those who want to hold up unguided chance. The cost of this single sonnet 10⁶⁹⁰ hopelessly exceeds the entire probabilistic budget of the universe (10¹⁴⁰). This calculation demonstrated to Flew that chance, operating within the finite bounds of the physical universe, cannot be the author of information with specified complexity.

Flew argued that this literary impossibility, as staggering as it is, is merely the beginning of the biological reality. The challenge for abiogenesis (the origin of life from non-life) is not just generating a random sequence of text, but generating a functionally specified sequence of chemicals. The biological “alphabet” consists of the 20 standard amino acids that form necessary proteins. A sonnet has specified complexity to convey meaning. A protein has specified complexity to perform a function. This function is determined by its precise three-dimensional folded shape.

• A single, medium-sized functional protein requires a specific sequence of perhaps 150 amino acids. The number of possible random sequences of this length is 20¹⁵⁰, which is approximately 10¹⁹⁵.

This added complexity makes the probabilistic chance argument profoundly more hopeless, as this sub-issue alone is also significantly larger than the total resources of the universe (10¹⁴⁰). Furthermore, the problem is that the vast majority of these 10¹⁹⁵ possible sequences are biologically useless. They will not fold into a stable, specific 3D shape that can perform a biological job, like acting as an enzyme or a structural component. The sequence must be almost exactly right. Therefore, the chance of a random chemical soup assembling even one functional protein is, for all practical purposes, zero.

This calculation, as impossible as it appears, still grants the materialist far too many assumptions. The 10¹⁹⁵ figure assumes that a pool of only the correct 20 amino acids is available, and that they will only link together in the correct way. This is not the reality. Amino acids generally exist in a mirror-image left-handed and right-handed form (a property called chirality). Life exclusively uses left-handed amino acids. A random soup would (if possible) make a roughly 50/50 mix. A single right-handed amino acid in the chain would act like a kink, destroying the protein’s 3D fold and its function.

• The odds of a 150-unit chain randomly selecting only left-handed components is 1/2¹⁵⁰, roughly 10⁴⁵.

Second, amino acids must be joined by a specific peptide bond. In a chemical soup, other, more likely bonds would form, creating a useless, non-protein structure. Therefore, to get one functional protein, a random process must first select only left-handed components (a 10⁴⁵ improbability), second link them all with only peptide bonds (another massive improbability), and third arrange them in the correct functional sequence (a 10¹⁹⁵ improbability). The total odds are not just added; they are multiplied, compounding a number of staggering impossibilities.

But as Flew points out, even this insurmountable hurdle is just the beginning. A single protein is not life, any more than a single gear is a Swiss watch. Life requires an integrated system. The simplest known free-living organism, Mycoplasma genitalium, requires 482 genes. These genes encode instructions for 482 different proteins, which must all work in a coordinated, interdependent factory. This is not just a collection of parts; it is an architecture where the failure of one component (like a repair enzyme or a power-supply protein) causes failure for the entire system. This entire system is encoded by its DNA, which Flew came to recognise as a true, symbolic information-bearing code. The DNA molecule is a digital, four-base (A, T, C, G) storage system. The sequence of bases, grouped in triplets called ‘codons’, functions symbolically. The codon ‘G-C-U’ means ‘Alanine’ in the same way ‘D-O-G’ means a canine. There is no physical or chemical reason for this; the relationship between the codon and the amino acid is arbitrary, like a language. Physics and chemistry can tell you no more about the information of the codon than they can tell you the meaning of the English language. This genetic code is a translation convention that is not dictated by the laws of chemistry. This presents an intractable chicken-and-egg problem. It is a closed loop that random chance cannot initiate. This is not a single paradox, but a complex series of them that must be solved simultaneously and immediately for life to exist:

1. Replication: To replicate the cell, the DNA must be copied. This requires an enzyme called DNA polymerase (a complex protein). But the instructions to build DNA polymerase are stored on the DNA.
2. Transcription: To make a protein, the DNA must be “unzipped” by helicase (a protein) and then “read” onto a messenger RNA (mRNA) molecule. This is done by RNA polymerase (another protein). The instructions for both helicase and RNA polymerase are on the DNA.
3. Translation: The mRNA is transported to a ribosome (a complex machine made of both proteins and rRNA). The ribosome reads the mRNA code. But the ribosome itself is built from parts whose instructions are on the DNA.
4. Decoding: The ribosome needs ‘translator’ molecules (tRNA) to fetch the correct amino acids. These tRNA molecules must be “charged” with the correct amino acid by a specific enzyme (aminoacyl-tRNA synthetases—themselves proteins). The instructions for these crucial enzymes are also on the DNA.
5. Energy: This entire process (replication, transcription, translation) is enormously energy-intensive. It requires a constant supply of ATP, the cell’s “fuel”. This ATP is generated by another set of complex protein machines (like ATP synthase), which are also encoded on the DNA.

This entire, irreducible system of DNA (information), proteins (machinery), and ATP (energy) must be present simultaneously for the system to function at all. You need the code to build the machines, the machines to read the code, and the energy to run the machines. A partial system is nothing and succumbs to the principles of entropy. The complexity of this genetic “text” and its decoding system is immeasurably greater than the entire collected works of Shakespeare.

Flew’s argument, then, became a simple “inference to the best explanation.” When faced with evidence, the best explanation is the one that accounts for all the evidence most appropriately and with the fewest contradictions. When we find a watch on the beach, we infer a watchmaker. When we read a sentence from Shakespeare, we rightly and immediately infer an intelligent author. We do this because the sequence is not random but functionally specified. It is an arrangement of parts that is incomprehensibly unlikely to occur by chance in an environment that actively ensures it won’t happen, and which is organised to achieve a specific purpose. Flew also pointed out the critical error in invoking natural selection as the “author.” Natural selection is proposed as a filter, but it is not a creator. It can only act on what already exists. It is a process of differential survival, and for survival to be “differential,” you must first have entities that are alive and self-replicating. Selection can explain the survival of the fittest, but not the arrival of the first. It cannot build the first self-replicating, information-rich molecule from the chemical soup of a pre-biotic earth. That origin event, which must precede any selection, remains an unreconcilable problem for materialism.

Flew concluded that the sheer, specified complexity of the information behind life pointed to God. The mathematical odds against a random, undirected process producing the DNA code and its associated cellular machinery are too great to be ignored. For Flew, the evidence did not support his long-held atheism. He was obligated by the principle of “following the evidence.” It must be noted, however, that it did not lead him to embrace any revealed religion. He described his new position as deism: the philosophical acceptance of a non-intervening, intelligent “First Cause” or “Divine Designer” in the classical, Aristotelian sense. He maintained that the origin of the information in DNA required an intelligent source, just as the origin of a Shakespearean sonnet requires an intelligent author. To deny this, in his view, was to abandon reason in favour of a prior commitment to materialism.