History of Artificial Intelligence: A Clear Timeline

The history of artificial intelligence stretches back more than 70 years, from early ideas about thinking machines to the chat assistants and image generators people use today. It has not been a straight climb. AI has gone through bursts of excitement, long periods of disappointment known as "AI winters", and sudden breakthroughs when new data, hardware and methods came together. This timeline covers the milestones that matter and what each one changed.

Before the 1950s: the idea of a thinking machine

Stories about artificial beings are ancient, but the scientific groundwork came in the first half of the 20th century. Mathematicians formalised logic and computation, and the first programmable electronic computers appeared during and after the Second World War. Once machines could follow stored instructions, the obvious next question was whether they could also reason.

1950: Alan Turing asks "Can machines think?"

In 1950 the British mathematician Alan Turing published "Computing Machinery and Intelligence". Instead of debating what thinking means, he proposed a practical test: if a person chatting by text cannot reliably tell a machine from a human, the machine could be said to show intelligent behaviour. The "Turing test" is still widely discussed, even though modern researchers see it as a limited measure.

1956: the Dartmouth workshop names the field

The term "artificial intelligence" became established at a summer research project at Dartmouth College in 1956, organised by John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon. The proposal argued that every aspect of learning could in principle be described precisely enough for a machine to simulate it. The Dartmouth workshop is generally treated as the founding moment of AI as a research field.

1950s to 1970s: early optimism

Early programs solved logic puzzles, played checkers and proved mathematical theorems. Researchers built simple neural networks such as the perceptron, and the chatbot ELIZA (1966) imitated a therapist by rephrasing what users typed. These systems impressed people, but they worked only in narrow, tidy settings. Computers were slow, memory was expensive and there was very little digital data to learn from.

The first AI winter

By the mid-1970s, progress had fallen far short of bold early predictions. Funding agencies in the US and UK cut support, and interest cooled. This period is often called the first AI winter. The lesson was that intelligence in the real world is much harder than solving well-defined puzzles.

1980s: expert systems

AI returned in the 1980s through expert systems, programs that encoded the knowledge of specialists as long lists of "if-then" rules. Businesses used them for tasks like configuring computer orders or diagnosing equipment faults. They delivered real value, but they were expensive to build, hard to update and brittle when they met situations their rules did not cover. When the market for specialised AI hardware collapsed in the late 1980s, a second AI winter followed.

1990s to 2000s: machine learning takes over

Researchers increasingly moved from hand-written rules to statistical machine learning, where systems learn patterns from data. In 1997 IBM's Deep Blue beat world chess champion Garry Kasparov, a symbolic milestone. Meanwhile, the internet produced huge amounts of text, images and user behaviour data, and machine learning quietly began powering spam filters, search ranking and product recommendations.

2010s: the deep learning breakthrough

Around 2012, deep neural networks trained on large image datasets using graphics processors (GPUs) dramatically improved image recognition. Similar gains followed in speech recognition and translation. In 2016 DeepMind's AlphaGo defeated top Go player Lee Sedol, a game long thought too complex for computers. In 2017 researchers at Google introduced the transformer architecture, which became the foundation for today's large language models.

2020s: generative AI goes mainstream

Large language models and image generators moved from research labs to everyday products in the early 2020s. Chat assistants made it possible to ask questions, draft text and write code in plain language, and image models could turn a sentence into a picture. Speech recognition and synthetic voices also became good enough for routine use. By 2026, multimodal models that handle text, images and audio together are common, and governments are putting rules in place, such as the EU AI Act.

Many of these capabilities are now packaged into simple tools. On TextPixAI, for example, the AI Chat assistant lets you try a modern language model without any setup.

What AI history teaches us

• Hype runs ahead of reality. Each wave brought predictions that did not come true on schedule.
• Data and computing power matter as much as ideas. Many methods existed for years before hardware made them practical.
• Narrow wins are real. Even in "winter" periods, useful AI kept spreading into everyday software.
• Trust has to be earned. As AI reaches more people, accuracy, fairness and accountability have become central concerns.

To track independent data on how the field is moving today, the Stanford AI Index publishes an annual report on research, investment and policy.

Key dates at a glance

• 1950: Turing publishes "Computing Machinery and Intelligence".
• 1956: the Dartmouth workshop gives the field its name.
• 1966: ELIZA shows that simple pattern matching can feel conversational.
• 1970s and late 1980s: two AI winters as funding falls.
• 1997: Deep Blue defeats Garry Kasparov at chess.
• 2012: deep learning transforms image recognition.
• 2016: AlphaGo beats Lee Sedol at Go.
• 2017: the transformer architecture is published.
• 2020s: generative AI reaches millions of everyday users, and the EU AI Act becomes law in 2024.

Related guides

Keep reading in this topic cluster:

• What Is Artificial Intelligence? A Plain-English Guide
• The Future of AI: What the Next 10 Years May Bring
• What Is the Technological Singularity? A Clear Guide
• What Is the Turing Test? Meaning, Limits and Today
• How Does Machine Learning Work? A Simple Explanation

Frequently asked questions

Who is called the father of AI?

John McCarthy is often given that title because he coined the term "artificial intelligence" and co-organised the 1956 Dartmouth workshop. Alan Turing is also widely credited for laying the theoretical groundwork.

What was the first AI program?

The Logic Theorist, created by Allen Newell, Herbert Simon and Cliff Shaw in the mid-1950s, is commonly cited as one of the first AI programs. It proved theorems from a mathematics textbook.

What were the AI winters?

They were periods, mainly in the 1970s and late 1980s, when funding and interest in AI dropped because results did not live up to expectations.

Why did AI suddenly improve in the 2010s?

Three things came together: much larger datasets, cheaper and faster GPUs, and better training methods for deep neural networks.