The story of artificial intelligence begins not with computers or robots, but with mathematics and logic. Between 1913 and 1950, a remarkable collection of mathematicians, logicians, and engineers established the theoretical foundations that would eventually make intelligent machines possible.
1910-1913: Principia Mathematica - The Logical Foundation
Authors: Alfred North Whitehead and Bertrand Russell
Publication: Cambridge University Press in three volumes (1910, 1912, 1913)
Details: This monumental 2,000+ page work attempted to derive all mathematical truths from a well-defined set of axioms and inference rules using purely symbolic logic. The notation introduced included symbols for logical operations (∧, ∨, ¬, →, ↔) that became standard in computer science. The work famously took 362 pages to prove that 1+1=2, demonstrating the rigor of formal logical systems. Though Gödel's incompleteness theorems later showed the impossibility of their goal, the symbolic manipulation techniques and formal reasoning methods became foundational to automated theorem proving, logic programming languages like Prolog, and the logical foundations of AI systems. The work directly influenced Church, Turing, and von Neumann, establishing the mathematical rigor that would characterize computer science.
1912-1915: El Ajedrecista - The First Chess Automaton
Creator: Leonardo Torres y Quevedo (Spanish civil engineer and mathematician)
First Demonstration: 1914 at University of Paris, refined version in 1920
Technical Details: The machine played the endgame of king and rook versus king using electrical sensors to detect piece positions on a special board with metallic pieces. It employed relay-based logic circuits, electromagnetic actuators to move pieces via magnets under the board, and a phonograph to announce "check" and "mate." The algorithm guaranteed checkmate within 63 moves from any legal position, implementing what would later be called a "state machine" with deterministic rules. Unlike Wolfgang von Kempelen's fraudulent "Mechanical Turk" (1770), this was genuine automation. Torres y Quevedo also invented the Telekino (1903), the first radio control system, and analytical engines, making him a pioneer in cybernetics and automation.
1920-1923: R.U.R. and the Birth of "Robot"
Author: Karel Čapek (Czech writer and playwright)
Play: "R.U.R. (Rossum's Universal Robots)"
Premiere: January 25, 1921, at National Theatre Prague
Etymology: The word "robot" came from Czech "robota" (forced labor/serf labor) and "robotník" (worker), suggested by Karel's brother Josef Čapek, a cubist painter. The play depicted artificial workers created through chemical/biological processes rather than mechanical construction, exploring themes of consciousness, rebellion, and humanity's relationship with artificial beings. Within two years, the play was translated into 30 languages and performed worldwide. The term "robot" immediately entered global vocabulary, replacing earlier terms like "automaton" or "mechanical man." The play's themes—robots developing emotions, demanding rights, ultimately rebelling—established narrative frameworks still used in AI fiction today, from Asimov's stories to "Westworld" and "Ex Machina."
1928-1936: Lambda Calculus - The Mathematics of Computation
Developer: Alonzo Church (Princeton University)
Key Publications: "A Set of Postulates for the Foundation of Logic" (1932-1933), "An Unsolvable Problem of Elementary Number Theory" (1936)
Technical Innovation: Lambda calculus introduced function abstraction (λx.M) and application, providing a formal system for expressing computation based on function definition, application, and variable binding. Church proved that lambda calculus was Turing-complete before Turing machines existed, showing it could compute any computable function. The system introduced concepts of bound/free variables, alpha-conversion (renaming), beta-reduction (function application), and eta-conversion (extensionality). This became the theoretical foundation for functional programming languages (LISP, Haskell, ML), type theory, proof assistants, and the mathematical semantics of programming languages. Church's students included Turing, Kleene, and Rosser, making Princeton the center of computability theory.
1931: Gödel's Incompleteness Theorems - The Limits of Logic
Author: Kurt Gödel (Institute for Advanced Study, Princeton)
Publication: "Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I"
First Theorem: Any consistent formal system containing arithmetic contains true statements unprovable within the system. Gödel constructed a statement essentially saying "This statement cannot be proven" using Gödel numbering to encode logical statements as numbers.
Second Theorem: No consistent system can prove its own consistency.
Impact on AI: These theorems demonstrated fundamental limitations of formal systems, showing that human mathematical intuition transcends mechanical proof procedures. This influenced debates about whether machines could fully replicate human intelligence, inspired Turing's work on undecidability, and established theoretical boundaries for what algorithms could achieve. The theorems suggested that any AI system based on formal logic would have inherent limitations, a crucial insight for understanding AI's theoretical constraints.
1936: The Turing Machine - Defining Computation
Author: Alan Mathison Turing (King's College, Cambridge)
Paper: "On Computable Numbers, with an Application to the Entscheidungsproblem"
Published: Proceedings of the London Mathematical Society, November 30 and December 23, 1936
Innovation: Turing introduced an abstract machine with an infinite tape divided into cells, a read/write head, a finite set of states, and transition rules. This provided the first precise mathematical definition of an algorithm and mechanical computation. Turing proved the Halting Problem was undecidable, showing no algorithm could determine whether arbitrary programs terminate. He demonstrated that a "Universal Turing Machine" could simulate any other Turing machine, prefiguring programmable computers. The Church-Turing thesis proposed that Turing machines capture all mechanical computation. This work established computability theory, influenced von Neumann architecture, and provided the theoretical foundation for computer science and AI.
1941: The Z3 - First Programmable Computer
Creator: Konrad Zuse (Berlin, Germany)
Completed: May 12, 1941
Specifications: Used 2,600 telephone relays, 22-bit word length (1 bit sign, 7 bit exponent, 14 bit mantissa), 5.3 Hz clock frequency, performed addition in 0.8 seconds and multiplication in 3 seconds. It consumed 4000 watts of power and weighed about 1 ton. The Z3 was programmed using punched 35mm film stock, implementing floating-point arithmetic and supporting loops (but not conditional branches). Destroyed in 1943 Allied bombing, it was reconstructed in 1961. Zuse also developed Plankalkül (1942-1945), the first high-level programming language, though not implemented until 1975. His work, isolated by WWII, independently developed many computing concepts, proving the feasibility of automatic computation essential for AI.
1943: The First Artificial Neural Network Model
Authors: Warren Sturgis McCulloch (neuropsychiatrist) and Walter Pitts (self-taught logician)
Paper: "A Logical Calculus of the Ideas Immanent in Nervous Activity" (Bulletin of Mathematical Biophysics)
Model Details: Proposed artificial neurons as binary threshold units with weighted inputs, demonstrating that networks could compute any Boolean function and were Turing-complete. Each neuron computed: output = 1 if Σ(weights × inputs) > threshold, else 0. They proved that cyclic networks could implement memory and sequential behavior. The paper merged neurophysiology, mathematical logic, and computation theory, inspiring von Neumann's computer designs and establishing the theoretical foundation for neural networks. Pitts was only 20 when co-authoring this foundational work, having run away from home at 15 to study with Russell at University of Chicago.
1944-1945: Game Theory and Stored-Program Architecture
Contributor: John von Neumann (Institute for Advanced Study, Princeton)
Game Theory (1944): "Theory of Games and Economic Behavior" with Oskar Morgenstern formalized strategic decision-making, introducing minimax theorem, zero-sum games, and expected utility theory. This provided mathematical frameworks for AI planning, multi-agent systems, and reinforcement learning.
EDVAC Report (June 30, 1945): "First Draft of a Report on the EDVAC" described stored-program architecture where instructions and data share memory, enabling self-modifying code and general-purpose computation. The "von Neumann architecture" featured a central processing unit, memory unit, and input/output, becoming the standard computer design. Von Neumann also contributed to cellular automata (self-replicating machines), the Monte Carlo method, and the merge sort algorithm.
1946: ENIAC - The Electronic Revolution
Creators: John Presper Eckert and John William Mauchly (University of Pennsylvania)
Public Unveiling: February 14, 1946
Specifications: 17,468 vacuum tubes, 7,200 crystal diodes, 1,500 relays, 70,000 resistors, 10,000 capacitors. Weighed 30 tons, occupied 1,800 square feet, consumed 150 kW of power. Could perform 5,000 additions or 357 multiplications per second—1,000 times faster than electromechanical machines. Programmed by setting switches and plugging cables, taking days to reprogram. The six primary programmers were women: Kay McNulty, Betty Jennings, Betty Snyder, Marlyn Meltzer, Fran Bilas, and Ruth Lichterman, whose contributions were largely unrecognized until recently. ENIAC computed artillery firing tables, hydrogen bomb calculations, and weather prediction, demonstrating electronic computation's potential for complex AI tasks.
1948: Cybernetics and Information Theory
Cybernetics - Norbert Wiener (MIT): "Cybernetics: Or Control and Communication in the Animal and the Machine" established the study of feedback systems, introducing concepts of homeostasis, feedback loops, and control systems. Wiener coined "cybernetics" from Greek "kybernetes" (steersman), developing mathematical theories of communication and control applicable to both machines and organisms. This influenced robotics, neural networks, and AI's understanding of intelligence as information processing.
Information Theory - Claude Shannon (Bell Labs): "A Mathematical Theory of Communication" created the mathematical foundation for the digital age. Shannon defined information as reduction in uncertainty (entropy), introduced the bit as the unit of information, proved the noisy channel coding theorem, and established limits on data compression and transmission. His framework enabled digital communication, data compression algorithms, and provided mathematical tools for measuring and optimizing information processing in AI systems.
1949: Learning Rules and Autonomous Robots
Hebbian Learning - Donald Olding Hebb (McGill University): "The Organization of Behavior: A Neuropsychological Theory" proposed that synaptic strength increases when neurons fire together, summarized as "cells that fire together, wire together." Mathematically: Δwij = η×xi×xj where wij is synaptic weight, xi and xj are neuron activations, and η is learning rate. This became fundamental to unsupervised learning, self-organizing maps, and competitive learning in neural networks.
Robotic Tortoises - William Grey Walter (Burden Neurological Institute, Bristol): Built "Machina Speculatrix" robots nicknamed Elmer and Elsie using two vacuum tubes, two sensors (light and touch), and two motors. These demonstrated complex behaviors from simple rules: seeking light when batteries were high, avoiding bright light and obstacles, and returning to recharge stations when batteries were low. This pioneered behavior-based robotics, subsumption architecture, and demonstrated emergence of complex behavior from simple rules.
1950: The Turing Test and Foundational AI Concepts
Turing Test - Alan Turing: "Computing Machinery and Intelligence" (Mind, October 1950) proposed the "Imitation Game" where a machine attempts to convince a human interrogator it is human through text conversation. Turing predicted that by 2000, machines with 128MB of memory could fool 30% of judges in 5-minute conversations. He addressed nine objections to machine intelligence, including theological, mathematical (citing Gödel), and consciousness arguments. The paper introduced concepts of machine learning, child machines, and the importance of education in AI development.
Computer Chess - Claude Shannon: "Programming a Computer for Playing Chess" (Philosophical Magazine, March 1950) outlined minimax algorithm with position evaluation, proposed two strategies: Type A (brute force) examining all moves to fixed depth, and Type B (selective) examining promising lines deeper. Estimated 10^120 possible chess games, introducing complexity analysis. Suggested evaluation functions considering material, mobility, king safety, and pawn structure.
Three Laws of Robotics - Isaac Asimov: Introduced in "Runaround" (1942), widely known by 1950:1. A robot may not injure a human or through inaction allow harm2. A robot must obey human orders except where conflicting with First Law 3. A robot must protect its existence except where conflicting with First or Second LawsLater added Zeroth Law: A robot may not harm humanity. These laws influenced AI ethics discussions and safety research.