When Was AI Created?

The Origin and Foundational Year of Artificial Intelligence

Artificial intelligence as a formal academic discipline was officially founded in 1956 at the Dartmouth Summer Research Project on Artificial Intelligence, held at Dartmouth College in Hanover, New Hampshire. Organized by computer scientist John McCarthy—alongside Marvin Minsky, Nathaniel Rochester, and Claude Shannon—the workshop introduced the term "artificial intelligence" and established the guiding hypothesis that every aspect of learning and intelligence could, in principle, be described so precisely that a machine could simulate it. Dartmouth workshop - Wikipedia Artificial Intelligence (AI) Coined at Dartmouth A PROPOSAL FOR THE DARTMOUTH SUMMER ...

While 1956 marks the official christening and institutional birth of the field, the conceptual and technical foundations of AI emerged across several distinct phases:

• 1943: Warren McCulloch and Walter Pitts published the first mathematical model of an artificial neural network, proving that networks of simplified artificial neurons could compute logical functions. The Dartmouth Conference, 1956: The Summer AI Was Born - isHistory
• 1950: British mathematician and cryptanalyst Alan Turing published the seminal paper "Computing Machinery and Intelligence," introducing the Imitation Game (now known as the Turing Test) to operationalize machine intelligence. What is the history of artificial intelligence (AI)?
• 1955–1956: Allen Newell, Herbert A. Simon, and Cliff Shaw developed the Logic Theorist, widely recognized as the first functional AI program, which proved mathematical theorems from Bertrand Russell and Alfred North Whitehead's Principia Mathematica. Dartmouth workshop - Wikipedia Artificial Intelligence (AI) Coined at Dartmouth
• 1956: The Dartmouth workshop formally established AI as a standalone research domain distinct from cybernetics, information theory, and automata studies. Dartmouth workshop - Wikipedia Artificial Intelligence (AI) Coined at Dartmouth

Chronological Evolution of AI Milestones

The history of artificial intelligence spans philosophical origins, formal mathematical foundations, algorithmic breakthroughs, and modern deep-learning paradigms.

Period / Year	Milestone / Breakthrough	Significance
1943	McCulloch-Pitts Neuron Model	First formalization of artificial neural networks using propositional logic. The Dartmouth Conference, 1956: The Summer AI Was Born - isHistory
1950	Turing's "Computing Machinery and Intelligence"	Shifted the philosophical inquiry from "Can machines think?" to empirical behavioral benchmarks. What is the history of artificial intelligence (AI)?
1956	Dartmouth Summer Research Project	Coined the phrase "artificial intelligence" and established research agendas for machine reasoning and search. Dartmouth workshop - Wikipedia Artificial Intelligence (AI) Coined at Dartmouth
1957–1958	Frank Rosenblatt invents the Perceptron	Early hardware-implemented artificial neural network capable of basic binary classification. Artificial Intelligence (AI) Coined at Dartmouth
1965	Joseph Weizenbaum creates ELIZA	Early natural language processing demonstration simulating a Rogerian psychotherapist via pattern matching. Artificial Intelligence (AI) Coined at Dartmouth
1974–1980	First "AI Winter"	Government and institutional funding collapsed due to unfulfilled performance promises and computational bottlenecks. Artificial Intelligence (AI) Coined at Dartmouth
1980s	Expert Systems & Backpropagation revival	Commercial adoption of rule-based systems (e.g., XCON) and resurgence of multilayer perceptrons via backpropagation. Artificial Intelligence (AI) Coined at Dartmouth
1987–1993	Second "AI Winter"	Specialized Lisp machine market crashed as general-purpose microprocessors overtook dedicated symbolic hardware. Artificial Intelligence (AI) Coined at Dartmouth
1997	IBM Deep Blue defeats Garry Kasparov	First computer system to defeat a reigning world chess champion in a tournament match under standard time controls. Artificial Intelligence (AI) Coined at Dartmouth
2012	AlexNet wins the ImageNet Challenge	Convolutional neural networks trained on GPUs demonstrated decisive performance improvements, igniting the modern deep learning era. Artificial Intelligence (AI) Coined at Dartmouth
2017–Present	Transformer Architecture & Large Generative Models	Self-attention mechanisms enabled scalable pre-training, resulting in contemporary large language models and multimodal systems. Artificial Intelligence (AI) Coined at Dartmouth

Early Conceptual and Mathematical Foundations (Pre-1956)

Before computer hardware could execute algorithmic logic, mathematical logicians established the theoretical boundaries of computation.

Turing Machines and Algorithmic Thought

In 1936, Alan Turing introduced the concept of the Universal Turing Machine—a theoretical device that manipulates symbols on a strip of tape according to a table of rules. Turing proved that such a machine could simulate any algorithmic computation, establishing that symbol manipulation forms the foundation of mechanical computation. In 1950, Turing published "Computing Machinery and Intelligence" in Mind, proposing that rather than debating subjective definitions of "thought," science should evaluate whether an automated system could generate natural language responses indistinguishable from a human interlocutor. What is the history of artificial intelligence (AI)? Artificial Intelligence (AI) Coined at Dartmouth

Cybernetics and Early Neural Modeling

Parallel to symbolic computation, biological cybernetics examined whether electrical networks could mirror human neurophysiology. Warren McCulloch and Walter Pitts demonstrated in 1943 that simple binary switching circuits could evaluate boolean expressions, demonstrating equivalence between formal propositional logic and idealized biological neural activity. In 1949, Donald Hebb formulated Hebbian learning theory, explaining that synaptic connections strengthen when connected neurons fire simultaneously ("neurons that fire together, wire together"), laying the theoretical groundwork for future machine learning algorithms. The Dartmouth Conference, 1956: The Summer AI Was Born - isHistory

The 1956 Dartmouth Workshop: Defining the Paradigm

The formal creation of AI as an academic discipline occurred in August 1956 during an eight-week workshop organized by John McCarthy at Dartmouth College.

                  +----------------------------------------------+
                  |  Dartmouth Summer Research Project (1956)   |
                  +----------------------------------------------+
                                         |
         +-------------------------------+-------------------------------+
         |                               |                               |
+------------------+           +-------------------+           +-------------------+
|  John McCarthy   |           |  Marvin Minsky    |           |  Allen Newell &   |
| (Term "AI", Lisp |           | (Neural networks, |           |  Herbert Simon    |
|   creator)       |           |  Minsky-Papert)   |           | (Logic Theorist)  |
+------------------+           +-------------------+           +-------------------+

The Founding Proposal

In 1955, McCarthy, Minsky, Rochester, and Shannon submitted a funding request to the Rockefeller Foundation titled "A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence". The proposal articulated the core thesis of symbolic AI:

"The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it." A PROPOSAL FOR THE DARTMOUTH SUMMER ...

The First Working Program: Logic Theorist

The primary technical breakthrough demonstrated during the 1956 workshop was the Logic Theorist, authored by Allen Newell, J.C. Shaw, and Herbert Simon of the RAND Corporation and Carnegie Institute of Technology. The program utilized heuristic search trees to automate the deduction of logical proofs. It proved 38 of the first 52 theorems in Chapter 2 of Principia Mathematica, finding a shorter, more elegant proof for one theorem than Russell and Whitehead had originally published. Dartmouth workshop - Wikipedia Artificial Intelligence (AI) Coined at Dartmouth

Major Eras of AI Development

Following its 1956 establishment, artificial intelligence progressed through alternating cycles of optimism, deployment, technical plateaus, and algorithmic reinvention.

1. The Symbolic & Heuristic Era (1956–1974)

Early AI focused on "Good Old-Fashioned AI" (GOFAI), relying on formal logic, search algorithms, and explicit symbolic representations. Notable achievements included:

• General Problem Solver (GPS): Developed by Newell and Simon (1959) to separate domain-specific knowledge from general-purpose problem-solving heuristics. Artificial Intelligence (AI) Coined at Dartmouth
• Lisp Programming Language: Created by John McCarthy in 1958, which became the primary language for AI development due to its native support for list manipulation and tree-structured symbol operations. Artificial Intelligence (AI) Coined at Dartmouth
• Micro-worlds and SHRDLU: Terry Winograd (1970) demonstrated natural language interaction with physical objects in a simulated 3D "blocks world." Artificial Intelligence (AI) Coined at Dartmouth

2. The AI Winters and the Rise of Expert Systems (1974–1993)

During the 1970s, funding agencies such as the U.S. Defense Advanced Research Projects Agency (DARPA) and the UK Science Research Council (via the 1973 Lighthill Report) sharply reduced AI investments due to scaling failures on real-world problems. Artificial Intelligence (AI) Coined at Dartmouth

The field rebounded in the 1980s through Expert Systems—knowledge-based programs designed to replicate human decision-making in narrow domains using thousands of "IF-THEN" rules (such as MYCIN for infectious blood diseases and XCON for computer configuration). However, the high maintenance costs and brittle failure modes of rule bases led to a second market correction between 1987 and 1993. Artificial Intelligence (AI) Coined at Dartmouth

3. Statistical Machine Learning and Deep Learning (1993–Present)

Beginning in the late 1990s, AI pivoted away from hand-crafted symbolic logic toward statistical inference, probability theory, and data-driven learning. Key structural transitions included:

• Empirical Validation and Benchmarking: Widespread adoption of standardized datasets (e.g., MNIST, ImageNet) established objective evaluation criteria. Artificial Intelligence (AI) Coined at Dartmouth
• Deep Neural Architectures: The convergence of backpropagation algorithms, massive parallel processing on Graphics Processing Units (GPUs), and broad internet datasets enabled multilayer neural networks to surpass human performance in image classification, board games (e.g., AlphaGo in 2016), and natural language processing via attention-based Transformer models. Artificial Intelligence (AI) Coined at Dartmouth

Symbolic AI vs. Connectionist AI: The Two Historical Paradigms

Throughout its evolution since 1956, AI research has been divided between two distinct technical philosophies regarding how machine intelligence should be constructed.

                           +-------------------------------+
                           | Artificial Intelligence (AI)  |
                           +-------------------------------+
                                          |
                +-------------------------+-------------------------+
                |                                                   |
    +-----------------------+                           +-----------------------+
    | Symbolic Paradigm     |                           | Connectionist / ML    |
    | (Top-Down / Logic)    |                           | (Bottom-Up / Data)    |
    +-----------------------+                           +-----------------------+
    | * Explicit Rules      |                           | * Neural Networks     |
    | * Knowledge Graphs    |                           | * Gradient Descent    |
    | * Deterministic Proof |                           | * Statistical Weight  |
    | * Expert Systems      |                           | * Deep Learning       |
    +-----------------------+                           +-----------------------+

1. Symbolic Paradigm (Top-Down): Posits that intelligence emerges from the manipulation of explicit symbols, concepts, and logical rules. It prioritizes interpretability and formal correctness but struggles with ambiguous, noisy real-world sensory data. Artificial Intelligence (AI) Coined at Dartmouth
2. Connectionist Paradigm (Bottom-Up): Posits that intelligence arises from networks of simple interconnected nodes that learn statistical weights directly from raw data via continuous mathematical optimization. While historically challenged by opacity ("black-box" models), connectionism forms the structural foundation of modern computer vision, speech recognition, and generative AI systems. The Dartmouth Conference, 1956: The Summer AI Was Born - isHistory Artificial Intelligence (AI) Coined at Dartmouth

The short answer: AI as a named field began in 1956

If you need a single year, the standard answer is 1956 — the year of the Dartmouth Summer Research Project on Artificial Intelligence, an eight-week gathering at Dartmouth College in Hanover, New Hampshire, that gave the field both its name and its founding research agenda. The term artificial intelligence itself is slightly older: it appears in the 1955 funding proposal written by John McCarthy, then a mathematics professor at Dartmouth, together with Marvin Minsky, Nathaniel Rochester, and Claude Shannon. Artificial Intelligence (AI) Coined at Dartmouth Dartmouth workshop A Proposal for the Dartmouth Summer Research Project on ...

So the most defensible formulation is: the phrase "artificial intelligence" was coined in 1955, the field was launched in 1956, and the underlying ideas and first working programs stretch from the 1930s to the late 1950s. Anyone who tells you AI was "created" in a single moment is compressing roughly two decades of mathematics, neuroscience, and early computing into a headline.

Why 1955–56 is the conventional answer

McCarthy's proposal asked for "a 2 month, 10 man study of artificial intelligence" to be carried out during the summer of 1956, and its famous conjecture set the field's ambition: that "every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it." The document, dated 1955 (sources give the date variously as late August or early September of that year), is generally credited with introducing the phrase into the technical vocabulary. Artificial Intelligence (AI) Coined at Dartmouth Dartmouth workshop A Proposal for the Dartmouth Summer Research Project on ...

McCarthy's choice of words was partly strategic. Related research already existed under labels such as cybernetics, automata studies, and complex information processing, and a new name helped mark out a distinct community with distinct goals rather than aligning the work with Norbert Wiener's cybernetics. The 1956 meeting mattered less for any single technical result — attendance was fluid and the group produced no unified breakthrough that summer — than for coalescing a set of people, problems, and shared vocabulary that became a durable discipline. Dartmouth workshop AI was born at a US summer camp 68 years ago. Here's ...

That is the key distinction to keep in mind: 1956 is when AI was founded as a field of study, not when intelligent machinery was first imagined, theorised, or built.

The prehistory: what existed before the name

The intellectual ingredients were assembled well before Dartmouth, and each strand answers a slightly different version of the question.

Formal logic and computation. The idea that reasoning could be reduced to symbol manipulation runs from Aristotle's syllogisms through Leibniz, George Boole's algebra of logic, and Gottlob Frege. Alan Turing's 1936 work on computable numbers supplied the crucial modern step: a precise mathematical model of what any mechanical procedure can do, which implies that a single universal machine can carry out any effective procedure — including, potentially, procedures we call thinking.

Artificial neurons. In 1943, at the University of Chicago, the neurophysiologist Warren McCulloch and the logician Walter Pitts published "A logical calculus of the ideas immanent in nervous activity," modelling the nervous system as a network of simple threshold elements that compute logical functions. These "McCulloch–Pitts neurons" are the ancestor of every artificial neural network in use today, which is why some accounts date the birth of AI — or at least of connectionist AI — to 1943 rather than 1956. A logical calculus of the ideas immanent in nervous activity A Logical Calculus of the Ideas Immanent in Nervous Activity Birthplace of Neural Networks: McCulloch & Pitts at UChicago

The philosophical framing. In 1950, Turing published "Computing Machinery and Intelligence" in the journal Mind, opening with the question "Can machines think?" and replacing it with an operational substitute he called the imitation game, now universally known as the Turing test. The paper also anticipated most of the standard objections to machine intelligence and sketched the idea of a "child machine" that learns rather than being programmed with adult knowledge. If you define AI's creation as the moment the goal was clearly stated, 1950 is a strong candidate. I.—COMPUTING MACHINERY AND INTELLIGENCE | Mind Turing test

Working machinery. Early digital computers of the late 1940s and early 1950s were already being used for tasks that looked like reasoning — game playing, theorem proving, and pattern recognition experiments — several years before the Dartmouth meeting. These programs were not called "AI" at the time because the label did not yet exist.

The first AI programs and the first learning machines

Immediately around the founding moment, the abstractions became running code. The Logic Theorist, developed by Allen Newell, Herbert Simon, and the programmer Cliff Shaw in 1955–56, could prove theorems from Whitehead and Russell's Principia Mathematica by searching a space of possible proofs with heuristics that pruned hopeless branches. It is frequently described as the first genuine AI program, and it established the "symbolic" or heuristic-search paradigm that dominated the field for its first three decades. history of artificial intelligence (AI)

Two related lines followed almost immediately:

• The perceptron. Frank Rosenblatt, working at Cornell Aeronautical Laboratory, described the perceptron in a 1958 paper and built physical implementations that learned to classify simple patterns by adjusting connection weights. It was the first widely publicised trainable neural network and the direct ancestor of modern deep learning, though its limitations — notably its inability to learn functions that are not linearly separable — contributed to a long period of neglect for neural approaches. Professor's perceptron paved the way for AI – 60 years too soon Perceptron
• Machine learning as a named idea. Arthur Samuel's checkers-playing program at IBM improved with experience, and his work in the late 1950s popularised the term machine learning for systems that acquire competence without being explicitly programmed for each case. Professor's perceptron paved the way for AI – 60 years too soon

By about 1960, then, everything the modern reader recognises as AI's basic repertoire existed in prototype: search, heuristics, knowledge representation, learning from data, and neural networks.

Why the "creation date" keeps moving

AI's history is not a smooth ascent but a sequence of enthusiasm cycles, and each cycle produces a new popular sense of when AI "arrived."

1. 1956–early 1970s — the founding optimism. Symbolic reasoning, general problem solvers, early natural-language systems, and confident predictions that human-level machine intelligence was perhaps a generation away.
2. Mid-1970s — the first "AI winter." Combinatorial explosion in search, critiques of perceptrons, and disappointing results in machine translation and speech led major funders in the United States and the United Kingdom to pull back.
3. 1980s — expert systems. Rule-based systems encoding specialists' knowledge found real commercial use, spawning a small industry, followed by a second retrenchment when the systems proved brittle and expensive to maintain.
4. 1990s–2000s — the statistical turn. AI shifted toward probability, statistics, and machine learning applied to data: speech recognition, search ranking, spam filtering, recommendation. Landmark demonstrations such as a computer defeating a reigning world chess champion in the late 1990s belong to this phase.
5. 2012 onward — deep learning. Large labelled datasets, graphics-processing hardware, and improved training methods made deep neural networks dramatically better at vision and speech, reviving the connectionist tradition that traced back to 1943 and 1958.
6. 2017 onward — transformers and generative AI. The Transformer architecture introduced in Google researchers' 2017 paper "Attention Is All You Need" replaced recurrence with self-attention and became the backbone of large language models; the public release of conversational chatbots built on such models in late 2022 is the point at which many people first experienced AI directly. What Is AI? Definitions and History - Matthew Sag

This is why the same question gets answers of 1943, 1950, 1956, 2012, or 2022 depending on who is asking. Each is a legitimate answer to a different question.

Choosing the right date for your question

The phrasing of the question usually reveals which milestone the asker wants:

• "What year was the term AI coined?" → 1955 (McCarthy's proposal), with the term entering wide use through the 1956 workshop.
• "When was AI created as a field?" → 1956, the Dartmouth Summer Research Project.
• "When was the first AI program made?" → 1955–56, the Logic Theorist, with earlier game-playing and theorem-proving programs as contenders depending on how strictly you define AI.
• "When were neural networks invented?" → 1943 for the mathematical model, 1958 for the first trainable, physically realised network.
• "When was the idea of thinking machines first proposed?" → Antiquity for the myths and automata; 1950 for the modern scientific formulation.
• "When was modern AI created?" → the 2012 deep-learning inflection and the 2017 Transformer, commercialised at scale from 2022.

Two caveats are worth stating plainly. First, "artificial intelligence" has never had a single agreed definition — it has covered symbolic reasoning, statistical learning, robotics, and pattern recognition at various times, and the boundary shifts as techniques mature and get renamed (once a capability becomes reliable, it tends to be called "software" rather than AI). Because the definition moves, the origin date moves with it. Second, priority claims in this history are genuinely contested: the Dartmouth participants themselves disagreed about what happened that summer, and different national research traditions — British cybernetics, Soviet automata theory, and American symbolic AI — have their own founding narratives. Dartmouth workshop AI was born at a US summer camp 68 years ago. Here's ...

For most purposes, the cleanest and best-supported sentence is this one: Artificial intelligence was named in 1955 and established as a research field at the Dartmouth workshop in the summer of 1956, building on work in logic, computing, and neuroscience from the 1930s and 1940s.