The History of Artificial Intelligence: From Ancient Philosophy to the Age of ChatGPT
- Stéphane Guy
- 2 days ago
- 10 min read
AI is now a hot topic all over the internet. But have you ever wondered how AI was created, what its origin(s) are... ? When we discuss artificial intelligence, we often think of computer tools like ChatGPT and Copilot, as well as AI-powered devices like Google Pixel smartphones or Galaxy S models. But was that really all there was to it in the beginning? Let us explain.
In Short
The history of artificial intelligence dates back to ancient philosophy, when Greek, Indian, and Chinese thinkers first attempted to systematise and mechanise human thought.
Before computing existed, early attempts to replicate living organisms, from Jacques de Vaucanson's mechanical duck to the myth of the golem, reflected a deep human drive to engineer artificial life and intelligence.
Modern AI was born in 1943 with McCulloch and Pitts' landmark neural network paper, then accelerated through milestones like the Turing Test and the 1956 Dartmouth Conference, where the term "artificial intelligence" was coined for the first time.
After a golden age in the 1950s–1960s, AI hit a severe funding drought in the 1970s before making a strong comeback in the 1980s thanks to advances in machine learning.
The 1990s proved AI's real-world viability with landmark wins like Deep Blue's defeat of Kasparov, paving the way for today's ubiquitous applications.
Before We Dive In: What Is Artificial Intelligence, Exactly?
Before tracing the arc of AI's history, it's worth anchoring ourselves to a solid definition. If you haven't already, take a look at our article What Is Artificial Intelligence?, it gives you the conceptual foundation you need to fully appreciate what follows.
AI's Origins: A Matter of Philosophy Before Computing
AI's origins are far less pragmatic than the sleek tools we use today. Artificial intelligence, in its deepest sense, stretches back thousands of years, to a time when Greek, Indian, and Chinese philosophers first tried to "mechanize" and systematize human thought, to break it apart, understand it, and ultimately replicate it. This philosophical drive, not mathematics or code, is where AI begins.
The tradition continued into the early modern period. Gottfried Wilhelm Leibniz, the 17th-century German polymath, laid out a compelling vision in his work De Arte Combinatoria, arguing that human thought could be organized according to a universal alphabet of shared ideas, a system that would allow any human reaction to be predicted and judged objectively.*
The implication was staggering: if human intelligence could be fully mapped, it could be replicated. Thousands of years before the first transistor, philosophers had already asked the question that still drives AI research today, how do you build an artificial mind if you don't yet understand your own?
AI and the Replication of Living Organisms
Running parallel to this philosophical tradition was a more tangible fascination: the mechanical replication of living creatures. The most famous example is Jacques de Vaucanson's mechanical duck, built around 1764, which was engineered to mimic a real duck's digestive system with uncanny precision.*
This wasn't mere spectacle. It reflected the era's obsession with understanding living systems from the inside out, not just biologically, but mentally. The same impulse appeared in religious and esoteric traditions across the world, from Tibetan monks to Jewish mystics who sought to create golems: beings endowed with autonomous consciousness, assembled entirely from raw materials, and programmed, if that word can be used, to obey their creators. The goal was precisely what we pursue today: an intelligence that is artificial yet fully functional.
1943–1956: The Theoretical Birth of AI
The AI we recognize today, computational, mathematical, reproducible, has a precise birth date: 1943.
That year, American neurologist Warren McCulloch and mathematician Walter Pitts published their landmark paper "A Logical Calculus of Ideas Immanent in Nervous Activity," presenting the scientific community with the first mathematical model of an artificial neural network. Their logic was elegant: if you want to build an artificial mind, start by modeling the human brain.*
By 1950, their work had inspired Marvin Minsky to build SNARC, the Stochastic Neural Analog Reinforcement Calculator, the world's first neural-network computer. SNARC was genuinely capable of learning from experience without being explicitly reprogrammed, a property that sounds almost mundane today but was revolutionary at the time.
In 1950, Alan Turing published his now-iconic proposal for what became the Turing Test: a benchmark designed to determine whether a machine is sophisticated enough to pass as human. The setup is simple. A human interrogator poses questions to an entity hidden behind a partition, not knowing whether it is human or machine. If the interrogator cannot reliably distinguish the machine from a person, the machine passes. Turing, one of history's greatest mathematicians, designed this not as a party trick but as a rigorous proxy for machine intelligence, a challenge that continues to shape AI research.
In 1952, Turing co-created Turochamp with statistician D.G. Champernowne, the first computer program designed to play chess. It was primitive, but it mattered. Then, in 1956, everything crystallised.
1956–1970: The Golden Age of AI
The Dartmouth Summer Research Project on Artificial Intelligence, held in the summer of 1956 at Dartmouth College in New Hampshire, is the founding moment of AI as a formal discipline. Organized by Marvin Minsky and mathematician John McCarthy, alongside Claude Shannon and Nathaniel Rochester, the conference brought together the field's brightest minds and, critically, gave the field its name: artificial intelligence was coined there for the first time.*
Three years later, in 1959, Arthur Samuel, IBM computer scientist and pioneer in computer science, coined the term "machine learning", defining it as giving computers the ability to learn without being explicitly programmed. The phrase stuck, and machine learning remains the cornerstone of the field today.
This era also produced heuristic search, one of AI's earliest operational methods, in which a program reaches its goal by using rules of thumb and evaluation functions to prioritize promising paths through a vast space of possibilities, rather than following a fixed script.
Between 1964 and 1966, MIT computer scientist Joseph Weizenbaum created ELIZA, the world's first chatbot, programmed to simulate a psychotherapist. Users were routinely stunned by how coherent and human its responses felt. Weizenbaum's own secretary reportedly asked him to leave the room so she could speak to the program in private.*
Nothing about today's chatbots is entirely new.
1970–1980: The AI Winter
By the early 1970s, the field that had generated such excitement was running into hard walls.
The first problem was technological: the hardware of the decade simply wasn't powerful enough to deliver on AI's promises. Researchers hit computational ceilings that no algorithm could compensate for, and the disillusionment spread fast, first through the scientific community, then through the investor class.
Academic criticism piled on simultaneously. Detractors challenged AI's lack of real-world applications, questioned the neural and mathematical models underpinning it, and raised ethical concerns that sound remarkably contemporary: Should we build a machine that thinks like a human? What does that mean for employment? For humanity's place in the world?
The combination, technological stagnation, academic skepticism, and ethical anxiety, triggered a sharp withdrawal of funding and institutional support. Historians of the field call this period the AI winter, a freeze that lasted roughly until the early 1980s. The risks that AI poses to society were being debated decades before most people had ever touched a computer.
1980–1990: AI's Return to the Research Agenda
AI thawed with the 1980s, driven largely by the practical application of a concept Arthur Samuel had theorized in 1959: machine learning.
This time, it wasn't just theory. AI systems were now connected to statistical methods and pattern-recognition techniques, enabling them to identify images and parse text in ways that felt qualitatively different from earlier programs. Transfer learning and iterative, error-based improvement became central to how these systems developed.
Investment followed. In 1981, Japan's Ministry of Economy, Trade and Industry committed 850 million dollars to its Fifth Generation Computer Systems project, an ambitious program aimed at building machines capable of natural conversation and human-like reasoning. The capital signal was unmistakable: AI was back, and serious money was chasing it.
1990–2000: AI's Great Victories
The 1990s mark the decade AI stopped being a laboratory curiosity and started generating real-world results, cautiously at first, then with growing confidence.
Applications emerged in finance and enterprise technology, though the public remained skeptical. A persistent gap between what AI could demonstrably do and what it was expected to do fueled frustration in some sectors. Ethical and philosophical resistance lingered. But regardless of perception, AI was threading itself into the economy.
Then came the moment that changed public consciousness entirely.
In May 1997, IBM's Deep Blue defeated world chess champion Garry Kasparov in a six-game match, the first time a computer had beaten a reigning world champion under standard tournament conditions.* The win wasn't just a chess result; it was a cultural inflection point, a demonstration that AI-powered computation could outperform the finest human strategic thinking in a domain considered uniquely human.
Separately, the DARPA Grand Challenge, run by the Defense Advanced Research Projects Agency, served as the proving ground for autonomous vehicles. In 2005, Stanford University's robot "Stanley", a modified Volkswagen Touareg, completed a 212-kilometer course through the Mojave Desert entirely without human input, navigating terrain it had never seen before and taking the $2 million prize.*
Both achievements were made possible by the same underlying force: exponentially faster processors capable of executing algorithms at scales that had been theoretically possible but computationally unthinkable a decade earlier.
2000–2020: AI's New Wins Against Humanity
The early 2000s gave AI its most powerful fuel source yet: the internet, and the incomprehensible volume of data it generated.
As the web expanded to every corner of the globe, billions of data points accumulated, text, images, behavior, preferences, creating what we now call Big Data. This dataset became, and remains, the primary training resource for modern AI. The more data, the smarter the model. And there had never been more data in human history. The environmental cost of processing all that data is a conversation the industry is only beginning to take seriously.
The major tech firms mobilized. Apple launched Siri in 2011, technically announced in 2007 but deployed publicly with the iPhone 4S. Google Now followed. In 2014, Microsoft introduced Cortana, inspired, famously, by the AI from the Halo video game franchise, and Amazon launched Alexa, still running today through its ecosystem of smart speakers.
Meanwhile, AI kept winning contests of intellect. IBM's win on Jeopardy! in 2011, a game requiring not brute calculation but the ability to parse ambiguous natural language clues and retrieve semantically precise answers under pressure. Then, in 2016, AlphaGo from Google DeepMind defeated world Go champion Lee Sedol, a feat many AI researchers had predicted was still a decade away. Go's near-infinite move space had been considered immune to the brute-force strategies that worked in chess.
2020 to Today: AI Democratized and Accessible to All
AI has since moved from a specialized tool to an ambient infrastructure.
The launch of ChatGPT in November 2022 marked a genuine inflection point: for the first time, a conversational AI capable of producing fluent, contextually rich text was available to anyone with a browser. That accessibility catalyzed a wave of launches, DALL-E for image generation, Microsoft Copilot for code and productivity, and an explosion of AI-native features across consumer hardware.
Today, the latest smartphone models do things that would have seemed science fiction five years ago. Google Pixel phones retouch images in seconds using on-device AI. Samsung Galaxy S series handsets translate conversations in real time. Recommendation algorithms on every major platform analyze listening and viewing behavior with enough precision to serve you content before you know you want it. The transformation of the music industry alone is a case study in how fast AI reshapes creative fields.
Professionally, the shifts are equally dramatic. Tools like Copilot accelerate software development, while text-to-image generators like DALL-E have restructured parts of the design industry. AI has also demonstrated meaningful progress in disability and medical research, opening access to services that were previously inaccessible to large populations. It's no longer a technology of the future. It's infrastructure, and it is reshaping how we work, create, and communicate in real time. The generative AI wave also brings invisible risks that deserve equal attention.
What's Next for AI? Futures and Stakes
The research community has rarely been more energized, or more divided.
Continuous advances are pushing toward what some experts describe as artificial general intelligence (AGI) or "strong AI": systems that don't just match human performance in narrow domains but exceed it across the board. The question of whether AGI is decades away or closer than we think is fiercely debated. What isn't debated is that despite nearly a century of progress, we are still in the early chapters of this technology's story.
The implications cut across every sector, medicine, education, transportation, communication, and the ethical stakes are rising in parallel. Questions about AI's impact on employment, the digital divide, and the outer boundary of what machines might one day become, including the contested concept of technological singularity, are no longer thought experiments. They're policy questions.
AI isn't just accelerating. It's compounding. And understanding where it came from is the clearest lens we have for understanding where it's going.
FAQ
When was artificial intelligence invented?
AI as a formal scientific discipline was born at the 1956 Dartmouth Summer Research Project, where the term "artificial intelligence" was coined for the first time. However, the conceptual roots stretch back to 1943, when McCulloch and Pitts published the first mathematical model of an artificial neural network, and further still to centuries of philosophical inquiry into the nature of human thought.
Who invented artificial intelligence?
No single person invented AI. The field emerged from a convergence of minds: Warren McCulloch and Walter Pitts modeled the first neural network (1943); Alan Turing proposed the Turing Test (1950); Marvin Minsky built SNARC, the first neural-network computer (1951); and John McCarthy organized the 1956 Dartmouth Conference that officially established AI as a research discipline. Arthur Samuel coined the term "machine learning" in 1959.
What was the first AI?
Depending on your definition, it was either SNARC (1951), the first neural-network computer, capable of self-learning, or ELIZA (1964–1966), the first program to hold a natural-language conversation with a human. The 1952 chess program Turochamp, co-developed by Alan Turing, also qualifies as an early AI system.
What is the AI winter?
The AI winter refers to a period of significantly reduced funding, research activity, and public interest in artificial intelligence, primarily during the 1970s. It resulted from a combination of technological stagnation (hardware wasn't powerful enough to support AI's ambitions), mounting academic criticism, and rising ethical concerns. A second, milder AI winter occurred in the late 1980s.
What was the most significant AI milestone of the 1990s?
Deep Blue's defeat of world chess champion Garry Kasparov in 1997 was the defining moment. It was the first time a computer system beat a reigning world champion under full tournament conditions, demonstrating that machine computation could surpass elite human strategic intelligence, a milestone that permanently shifted public and scientific perception of what AI could achieve.
What is machine learning and when was it invented?
Machine learning is the field of AI concerned with systems that learn from experience rather than being explicitly programmed for every scenario. The term was coined by Arthur Samuel in 1959. Today, it underpins virtually every AI application, from recommendation engines to large language models like ChatGPT. You can explore the full breakdown in our guide to supervised, unsupervised, and reinforcement learning.
Could AI ever surpass human intelligence entirely?
This remains one of the most contested questions in both AI research and philosophy. Some experts believe artificial general intelligence, a system exceeding human performance across all cognitive domains, is achievable within decades. Others argue it remains structurally out of reach. The concept of a technological singularity, a hypothetical point at which AI becomes self-improving at an uncontrollable rate, sits at the center of this debate.