Colendi AIResearch
Menu

Research archive

The Intelligence Papers

A navigable record of the ideas, artifacts, and institutions that formed the modern intelligence stack—reviewed as evidence, not arranged as a reading list.

211reviewed records

22research domains

1843–2025year range

19 records

0011936Origins & Computability

Foundational paper

On Computable Numbers (Turing Machine)

Alan Turing

Turing described a simple imagined device that reads and writes symbols on a tape according to a finite table of rules, and argued this captures anything a human clerk could compute by following steps. Using it he showed that a single 'universal' machine can simulate any other by reading its description, and that the halting problem has no general algorithmic solution. This gave a rigorous definition of 'algorithm' and 'computable', and the universal-machine idea underlies the concept of a programmable general-purpose computer.

UnknownDifficulty 6/10Verified
Read record Primary source
0021945Origins & Computability

Architecture report

First Draft of a Report on the EDVAC (stored-program architecture)

von Neumann

The draft described a design with a central arithmetic unit, a control unit, and a shared memory holding both program and data, communicating over common paths. Because instructions live in modifiable memory, the same hardware can run any program just by loading different contents. This 'stored-program' organization became the template for essentially all subsequent general-purpose computers.

UnknownDifficulty 5/10Verified
Read record Primary source
0031948Origins & Computability

Foundational paper

A Mathematical Theory of Communication (information theory)

Claude Shannon

He measured information in bits using entropy, separating a message's content from its meaning, and modeled communication as a source, channel, and receiver subject to noise. He proved that every channel has a maximum reliable rate (its capacity) and that codes exist to approach it with arbitrarily few errors, and gave limits on lossless compression. These results underpin modern data compression, error-correcting codes, and digital communication and storage.

UnknownDifficulty 6/10Verified
Read record Primary source
0041937Origins & Computability

Thesis

A Symbolic Analysis of Relay and Switching Circuits

Claude Shannon

In his master's thesis Shannon mapped open and closed switches onto Boolean true/false values and series/parallel wiring onto logical operations. This meant a designer could write a switching circuit as a logic equation, reduce it algebraically to use fewer components, and verify it behaves correctly. The result became the standard method for designing the digital logic inside telephone systems and later computers.

UnknownDifficulty 5/10Verified
Read record Primary source
0051943Origins & Computability

Foundational paper

A Logical Calculus of Ideas Immanent in Nervous Activity (artificial neuron)

Warren McCulloch, Walter Pitts

They abstracted a neuron as a device that sums weighted inputs and fires if the total crosses a threshold, ignoring biological detail. They proved that networks of such idealized neurons can implement any logical proposition, linking brain-style computation to formal logic and Turing's model. This established the idea that networks of simple connected units can compute, providing the conceptual starting point for later artificial neural networks.

UnknownDifficulty 5/10Verified
Read record Primary source
0061945Origins & Computability

Vision essay

As We May Think (Memex) + Science the Endless Frontier

Vannevar Bush

Bush described a hypothetical machine that would hold a person's books, records, and notes on microfilm and let the user create named 'trails' of associative links between items. He argued that organizing knowledge by association rather than rigid indexing would match how people actually think and recall. The essay is widely credited as an early inspiration for hypertext, personal information systems, and the linked structure of the web.

UnknownDifficulty 4/10Verified
Read record Primary source
0071950Origins & Computability

Foundational paper

Computing Machinery and Intelligence (Turing Test)

Alan Turing

Turing proposed the 'imitation game', in which a judge exchanges typed messages with a hidden human and a hidden machine and tries to tell which is which. He suggested that a machine passing this test should count as intelligent for practical purposes, sidestepping arguments over the definition of thought. The paper also anticipated and answered common objections and outlined machine learning, setting an early behavioral benchmark and framing debates in artificial intelligence.

UnknownDifficulty 4/10Verified
Read record Primary source
0081945Origins & Computability

Institutional milestone

The Manhattan Project / Los Alamos (big-science template)

J. Robert Oppenheimer, Leslie Groves, John von Neumann, et al.

Between 1942 and 1945 the United States gathered physicists, chemists, and engineers at Los Alamos and other sites to build an atomic weapon, coordinating theory, experiment, and mass manufacturing on an unprecedented scale. The effort showed that a well-funded, centrally managed team could compress decades of research into a few years. It created the postwar model of the national laboratory and the pattern of governments financing ambitious technical projects. Many figures and methods from this work, including early electronic computation, carried directly into the founding of modern computer science.

UnknownDifficulty 3/10Verified
Read record Primary source
0091949Origins & Computability

Method

The Monte Carlo Method / Metropolis Algorithm

Stanislaw Ulam, John von Neumann, Nicholas Metropolis

Working on nuclear physics problems at Los Alamos, Stanislaw Ulam, John von Neumann, Nicholas Metropolis, and colleagues realized that random sampling on a computer could approximate answers to equations too complex to solve directly. The 1953 Metropolis algorithm added a rule for sampling from a probability distribution by accepting or rejecting proposed moves, making it possible to simulate systems in equilibrium. These techniques became standard tools across physics, statistics, and later machine learning. Much of modern probabilistic modeling and Bayesian computation traces back to this sampling idea.

UnknownDifficulty 6/10Verified
Read record Primary source
0101956Origins & Computability

Founding event

The Dartmouth Workshop (birth of Artificial Intelligence)

John McCarthy, Marvin Minsky, Claude Shannon, Nathaniel Rochester

In the summer of 1956, John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon organized a workshop at Dartmouth College to study whether machines could be made to reason, learn, and use language. Their proposal coined the term 'artificial intelligence' and argued that every aspect of intelligence could in principle be specified well enough to be automated. The meeting gathered the researchers who would lead the field for decades. It set the research directions, from problem solving to language and learning, that defined AI's first generation.

UnknownDifficulty 3/10Verified
Read record Primary source
0111958Origins & Computability

Foundational paper

The Perceptron

Frank Rosenblatt

In 1958 Frank Rosenblatt described the perceptron, a simple network that combines weighted inputs and produces a decision, and a procedure that updates those weights whenever it makes a mistake. This gave a working example of a machine that improves its performance by seeing labeled data rather than being explicitly programmed. The perceptron could learn to separate patterns that are linearly separable, though later work showed its limits on harder problems. Its learning rule and layered structure are the foundation of today's deep learning.

UnknownDifficulty 5/10Verified
Read record Primary source
0121854Origins & Computability

Foundational treatise

Boolean Algebra (The Laws of Thought)

George Boole

Boole treated statements as variables taking values that behave like 0 and 1, with operations for 'and', 'or', and 'not' following consistent algebraic rules. This turned deductive reasoning into symbol manipulation that could be checked mechanically rather than by intuition. The system later became the mathematical basis for digital logic circuits and for how computers represent and evaluate logical conditions.

UnknownDifficulty 4/10Verified
Read record Primary source
0131843Origins & Computability

Foundational design

Analytical Engine & the First Algorithm

Charles Babbage, Ada Lovelace

The Analytical Engine was a proposed mechanical calculator with a separate store and processing 'mill', able to follow instructions and branch conditionally rather than compute one fixed formula. In her annotations to a paper describing it, Lovelace wrote out a step-by-step procedure for computing Bernoulli numbers, often cited as the first published algorithm intended for a machine. She also argued the engine could operate on symbols beyond numbers, anticipating general-purpose computation, though the machine was never built.

UnknownDifficulty 4/10Verified
Read record Primary source
0141948Origins & Computability

Foundational book

Cybernetics: Control and Communication in the Animal and the Machine

Norbert Wiener

Cybernetics framed systems as maintaining goals by sensing their output and feeding it back to adjust their input, whether the system is a thermostat, an animal, or a servomechanism. Wiener drew together control theory, feedback, and communication to treat purpose and self-correction as engineering problems. The framework influenced control engineering, early thinking about learning machines, and how researchers modeled adaptive and self-regulating behavior.

UnknownDifficulty 5/10Verified
Read record Primary source
0151949Origins & Computability

Foundational book

The Organization of Behavior (Hebbian learning)

Donald Hebb

Hebb argued that when one neuron persistently helps fire another, the link between them grows stronger, often summarized as 'cells that fire together wire together'. He used this to explain how groups of neurons form 'cell assemblies' that represent learned concepts and associations. The rule gave neuroscience a concrete account of learning and became a foundational principle for training weights in artificial neural networks.

UnknownDifficulty 4/10Verified
Read record Primary source
0161943Origins & Computability

Wartime engineering

Enigma / Bletchley Park codebreaking (Bombe / Colossus)

Alan Turing, Gordon Welchman, Tommy Flowers, Bill Tutte

During the Second World War, British codebreakers built machines to break the German Enigma and Lorenz ciphers by mechanizing the search through vast numbers of possible settings. The Bombe automated the logical elimination of impossible key configurations, while Colossus used electronic valves to process teleprinter traffic at high speed. This work demonstrated that computation could replace human labor on structured reasoning tasks. It also produced engineering experience and people, including Alan Turing, who shaped the first general-purpose computers.

UnknownDifficulty 5/10Verified
Read record Primary source
0171944Origins & Computability

Foundational book

Theory of Games and Economic Behavior (game theory)

John von Neumann, Oskar Morgenstern

In 1944 John von Neumann and Oskar Morgenstern set out a mathematical theory of how rational agents should act when their outcomes depend on each other's choices. They formalized games, strategies, and payoffs, and proved results such as the minimax theorem for zero-sum games, along with a theory of utility for decisions under uncertainty. The book gave economics and later computer science a rigorous way to reason about competition and cooperation. Its concepts underpin work on reinforcement learning, mechanism design, and multi-agent AI.

UnknownDifficulty 6/10Verified
Read record Primary source
0181964Origins & Computability

Foundational paper

Algorithmic Information Theory (Solomonoff / Kolmogorov)

Ray Solomonoff, Andrey Kolmogorov, Gregory Chaitin

In the 1960s Ray Solomonoff, Andrey Kolmogorov, and Gregory Chaitin independently proposed measuring how complex a string of data is by the size of the smallest computer program that can generate it. A string that needs a long program is effectively random, while a compressible one is simple. Solomonoff used this idea to build a theory of prediction that favors the simplest explanation consistent with the data. These concepts connect compression, probability, and learning, and they inform how researchers reason about generalization and Occam's razor in machine learning.

UnknownDifficulty 7/10Verified
Read record Primary source
0191969Origins & Computability

Critique

Perceptrons (Minsky-Papert critique)

Marvin Minsky, Seymour Papert

Through geometric and algebraic analysis the authors characterized exactly which predicates a single-layer perceptron can and cannot represent, showing that some require information the local, limited-order perceptron cannot combine, with XOR and connectedness as prominent examples of failures. They noted that multilayer networks could in principle overcome these limits but that no effective training method for them was then known. The rigorous negative results are widely credited with cooling early enthusiasm and funding for neural network research until multilayer training via backpropagation was popularized in the 1980s.

UnknownDifficulty 5/10Verified
Read record Primary source