Foundations of AI: Creativity, McCulloch and Pitts, and the Fusion of Neurons and Mathematics
Creativity is often misunderstood as creating something new from nothing when it is fundamentally about seeing the familiar in an unfamiliar way. Whether it is a scientific breakthrough, artistic revolution, the sense of the comic, or spiritual transformation, this truth holds good.
When a serious scientist or an artist becomes so involved in solving a problem that it saturates their body and mind, then, under certain conditions (not entirely within the control of an individual), new connections bubble up from the depths of the subconscious, casting a completely different light on the problem and resolving an insurmountable knot. Arthur Koestler's magnificent study of creativity, The Act of Creation, introduced the word "bisociation" to indicate this condition - how different frames of reference or patterns can occasionally collide or combine to provide fresh insights into solving an artistic cul-de-sac or resolving a scientific conundrum.
We label such moments as "Eureka" moments, and we also incorrectly glorify these seminal moments as happening without effort. Both these positions are inaccurate. Before a breakthrough, countless hours of sustained effort working on the problem precede the moment when such unexpected connections arise, either through a chance exposure to some other frame of reference or through a clarifying insight welling up from within. Either way, it is this fresh perspective that helps resolve the problem. History books omit the period of struggle and only project moments of breakthrough. The current AI revolution has witnessed several breakthroughs stemming from interdisciplinary connections. However, none exemplifies the power of creativity better than the story of how the workings of neurons in the brain became the foundation for modeling machine computation and artificial intelligence.
The idea that machine learning can be modeled like the neurons in the brain is a breathtaking convergence of two separate frames of reference: one biological and the other mathematical. For centuries, the brain was a mystery (it still is). Then, in the late 19th century, Santiago Ramón y Cajal, the great Spanish neuroscientist, peered through a microscope and saw something never seen before: a specific type of cell - the neuron. Cajal's work revealed that the brain was not a tangled mass, as once believed, but a network of distinct, individual neurons creating an "enchanting loom," in the immortal words of Sir Charles Sherrington. Each neuron received inputs at one end and fired a signal at the other. Our thoughts, imagination, consciousness, and perceptions were all somehow related to the firing of these neurons. Still, how the dance of these billions of neurons conjured the inner self and the isomorphic reflections of the world it navigated remained a mystery. Cajal's discovery won the Nobel Prize in Physiology or Medicine in 1906 and set the tone for the rapid advancement of neuroscience as a scientific discipline. From the mathematical side, the goal of Russell and Whitehead's Principia Mathematica was to reduce the whole of mathematics to its fundamental propositions. These propositions would either lead to "true" or "false"- in other words, a binary decision. From a series of such binary choices built on top of one another, Principia attempted to make the edifice of mathematics.
It took two brilliant minds - one a seasoned philosopher and neurophysiology student, the other a mathematically gifted runaway - to come together under extraordinary circumstances and reshape our understanding of the brain. While one man was 42 years old, a philosopher, psychologist, and neurophysiology student, the other was a runaway from home, just 18 years of age, mathematically brilliant but without any academic credentials. The story of Warren McCulloch and Walter Pitts’s collaboration, friendship, and breakthroughs would change the landscape of information technology forever. Between them, they would create the first mechanistic theory of the mind, the first computational theory of neuroscience, the logical design of modern computers, and lay the foundations of AI, which is sweeping the world today. It is important to know both these men and how they changed the world. Let us start with Walter Pitts.
Walter Pitts was a Detroit boy who had self-educated himself in Greek, Latin, logic, and mathematics. His father never hesitated to use his fists to discourage Pitts from going to school, preferring that he work instead. By nature, Pitts was a timid boy, and his shy, introspective demeanor was ready fodder for relentless bullying by his classmates. The young boy found no empathy for his prodigious talent within his family or at school. The library was the only place he felt at home, his refuge and a sanctuary of silence and knowledge where he could be left alone to read and think.
One day, on a cold afternoon when Pitts was twelve years old, he was on the run from his bullies. He sprinted to the library and hid among its vaulted shelves. As darkness fell, his pursuers had long since forgotten him, but Pitts, still frightened, had no intention of leaving. His eyes fell on a three-volume tome: Principia Mathematica, Bertrand Russell, and Alfred Whitehead’s monumental study of logic. Published between 1910 and 1913, it attempted to reduce all mathematics to logic - its first principles. For even an experienced mathematician, Principia is an impenetrable labyrinth of symbols and cryptic statements. It is famously said that the proof for 1+1=2 appears on page 359 of the book. But for Pitts, The Principia was not an obstacle; it was an invitation. He immersed himself in its pages. For three unbroken days, he remained in the library until he had read all three volumes. Not only had he read the book, but he also discovered logical flaws in Russell's reasoning. With the audacity only a youngster could muster, he composed a letter to the great philosopher, pointing them out with razor-sharp precision.
Russell, always a gracious teacher, read the letter and was astonished by the critique. He wrote back, inviting Pitts to Cambridge to study and work with him. Obviously, Pitts couldn’t go -he was only twelve years old. Three years later, when Pitts was fifteen, Russell visited Chicago for a conference at the University of Chicago. Upon hearing that Russell was in America, Pitts decided it was time to meet him. He left home in the middle of the night without telling anyone (never again in his tragically brief life to see any member of his family or speak of them). Though his flight from home led him to his destiny, in a personal sense, it left him without any roots. The only person he would trust and consider family for the rest of his life was unknowingly awaiting him in Chicago. Life, when seen in retrospect, was playing her choicest cards, bringing together two people from different backgrounds, educations, and upbringings to spark an intellectual collaboration that would usher in the age of AI.
The irony is that Walter Pitts never met Bertrand Russell. He hung around the university corridors doing odd jobs to sustain himself and occasionally sneaked into lecture rooms to listen to the professors. A young medical student, Jerome Lettvin, whom he befriended, introduced Pitts to Warren McCulloch, a middle-aged professor with an excellent reputation who had just moved from Yale to the University of Chicago. That introduction would be the inflection point for a new age in human history.
McCulloch was a confident, gray-eyed, wild-bearded, chain-smoking philosopher-poet who thrived on whiskey and ice cream and never went to bed before 4 a.m. His stable and prosperous home with his wife, Ruth, and their children was a sanctuary for intellectuals and literary minds of Chicago, who often gathered late into the night, passionately debating everything under the sun - from philosophy to literature to the Spanish Civil War that raged at the time. Though a mathematician, philosopher, and trained neurophysiologist, McCulloch was, at his core, a philosopher wrestling with the profound enigma of consciousness and the basis of thought. Was thought, as Descartes believed, different from matter? Or was it a product of matter, in which case there must be a basis for it? This was the problem that McCulloch was grappling with. McCulloch admired Gottfried Leibniz's work - the 17th-century polymath and Newton’s contemporary. Leibniz had daringly speculated about a system akin to an English alphabet, where each letter symbolized a concept, and by applying logical rules to these symbols, all human thought could be codified. McCulloch's understanding of the neuron led him to believe that there should be a mathematical model for how the neurons process information. Turing’s groundbreaking paper (just published) suggested that such an approach was not only conceivable but potentially realizable.
McCulloch found in Pitts a remarkable capacity for rigorous logical thought and an ability to sift through the periphery of any complex problem and get to its core. When McCulloch explained to Pitts that he was trying to model a mechanical computational process using Leibniz’s idea of an alphabet soup and the binary methods of Russell’s Principia, Pitts was electrified. After all, both men loved and had digested Principia. They found a meaningful resonance in the conception of a single neuron as a logical gate - a proposition in Russell's language that could output a binary decision: a “yes” or a “no,” an electrical signal passing through or not, based on the strength of the inputs and the logical operations performed on them.
McCulloch struggled to find the proper mathematical framework to represent the model of a neuron. Leibniz’s logical calculus, which he initially tried using, involved time, and he repeatedly ran into the problem of his mathematical model forming loops - an output from one neuron becoming an input to a previous neuron in the chain. This was mathematically impossible using traditional calculus, where time played a crucial role. Pitts knew the solution: Modulo mathematics, a branch of mathematics that worked with numbers that cycled like the hours on a clock.
Every night, after dinner, McCulloch and Pitts would spread their papers on the dining table and work feverishly throughout the night to model a mechanical neuron. Rook, McCulloch’s wife, would quietly clean up and take their daughters to bed -except the eldest, who would sometimes sit around the table, listen to the discussions, and sketch the neuronal representations her father and Pitts debated. Her diagrams would become part of their final paper.
Pitts mathematically convinced McCulloch that neurons looping back on themselves were inevitable in the brain. A loop that closed on itself became memories and conceptual abstractions, while those that continued down the chain were the computational equivalent of data processing: memory and information processing - two key aspects of the brain. In retrospect, this insight seems simple, but at the time, it was revolutionary. No one had yet connected Boolean algebra, Cajal’s neuron theory, Russell’s Principia, and Modulo mathematics. All these domains had to click together to represent the mechanical equivalent of a neuron. It was still a model, a way of looking at the brain - just as Niels Bohr’s model of the atom was a way of making sense of atomic behavior. In both cases, the models validated empirical results, and that is what science is all about.
The defining moment of their collaboration came in 1943 when McCulloch and Pitts co-authored a paper that would change the course of modern science: A Logical Calculus of the Ideas Immanent in Nervous Activity. This paper proposed the first artificial neural network model - a mathematical framework that demonstrated how neurons could perform logical operations akin to a Turing machine. They showed that the brain could, in principle (this distinction is important; it was a model that seemed to work), be understood as a logical system where neurons operated like switches, turning on or off based on inputs. It provided the blueprint for artificial neural networks, the very foundation of modern machine learning and artificial intelligence.
Pitts found in McCulloch everything he missed during his formative years - unconditional acceptance, deep friendship, intellectual companionship, and the father figure he never had. Although he lived with the McCullochs only for a short time, Pitt would always refer to McCulloch’s house as his only home for the rest of his life. For his part, McCulloch had found in Pitt a kindred spirit, a razor-sharp mind that brought McCulloch’s notions to life, a son perhaps. As he put it in a letter of reference about Pitts, “Would I have him with me always?" Despite Pitt's intellectual contributions with McCulloch, Norbert Wiener, and Von Neumann over the next few decades, he remained a deeply troubled individual. Years of childhood trauma and emotional isolation had left lasting scars. While McCulloch provided embracing companionship whenever and however he could, Pitts struggled with holding on to personal relationships and never fully integrated anywhere. His stature as a thinker and a logician was legendary. Von Neumann would often say that if Pitts had reviewed a paper, then nothing more needs to be done. Such was his intellectual integrity. In later life, Pitts slipped into alcoholism, which further deepened his reclusive nature.
In 1969, Pitts and McCulloch died within months of each other - one from cirrhosis of the liver and the other from a heart attack. Pitts was only 46 years old when he died. The number of breakthroughs he ushered in during his brief existence was nothing short of incredible. The intellectual umbilical cord between McCulloch and Pitts was never severed. They had their differences and took a different path after their initial collaboration, but they continued to sustain and collaborate till the end.
Their legacy endures. Generative AI owes its origins to these two men who brought an interdisciplinary focus to lay the foundation of computing. Every time an AI model decides, every time a neural network recognizes a face, translates a language, or mimics human cognition, it stands on the foundation built by a brilliant, self-taught teenager who once sought refuge in the silent aisles of a library, and a middle-aged philosopher who dared to mimic the brain.
Author’s bio: Balasubramaniam N. is an Associate Vice President at NIIT’s Center of Excellence (COE), bringing over 25 years of expertise in Consulting, Training, and architecting on NoSQL solutions and Data science.