The History of AI

From vision to reality

The history of artificial intelligence is a fascinating journey full of breakthroughs, setbacks, and unexpected turns. What was once considered science fiction is now part of our everyday lives. To understand where we stand today, it is worth looking back at the most important milestones.

The Turing test (1950): In 1950, the British mathematician Alan Turing published his famous essay "Computing Machinery and Intelligence". In it, he asked the provocative question: "Can machines think?" He proposed an experiment, known today as the Turing test, in which a human communicates via text messages with a machine and with another human. If the questioner cannot tell which is the human and which is the machine, the machine has passed the test. This test still shapes AI research today.

The term "artificial intelligence" was officially coined in 1956 at the Dartmouth Conference. Researchers such as John McCarthy, Marvin Minsky, and Claude Shannon were convinced that machines would reach human intelligence within a generation. This enthusiasm led to the first programs that could play chess, prove mathematical theorems, or simulate simple conversations.

The ups and downs of AI research

The pioneering era and the first AI winter

In the 1950s and 1960s, there was great euphoria. Programs such as ELIZA (1966) simulated simple therapy conversations, and research made rapid progress in logical problem solving. But the initial promises were too ambitious. When it became clear that real human intelligence is far more complex than expected, governments and companies drastically cut their funding. The first AI winter (approx. 1974–1980) had begun, a phase of disillusionment and reduced research.

Expert systems and the second AI winter

In the 1980s, AI made a comeback thanks to expert systems. These programs encoded the knowledge of human experts as rules ("If symptom X and Y, then diagnosis Z"). Companies invested billions. But expert systems were expensive to maintain, brittle with unexpected input, and hard to update. The second AI winter (approx. 1987–1993) followed. A highlight of the 1990s: in 1997, IBM's Deep Blue defeated the reigning world chess champion Garry Kasparov, a symbolic milestone, even though Deep Blue worked more with raw computing power than with "intelligence".

Big data and deep learning

With the internet, the amount of available data exploded. The turning point came in 2012: the neural network AlexNet won the ImageNet competition by a margin that shook the experts. Deep learning, the training of deep neural networks, suddenly became practical, thanks to powerful GPUs and huge datasets. In 2016, Google's AlphaGo defeated Lee Sedol, a multiple world Go champion. Go had been considered too complex for computers. This victory came as a shock. In 2017, Google researchers published the paper "Attention Is All You Need" and introduced the Transformer architecture, the foundation for everything that came after.

The Transformer revolution and generative AI

From 2018, language models such as BERT and GPT fundamentally changed the landscape. According to a UBS estimate, ChatGPT (November 2022) reached around 100 million active users within just two months, making it one of the fastest-growing applications ever. Suddenly anyone could "talk" to AI. Since then, the pace has accelerated further: multimodal models understand text, images, audio, and video at the same time. AI agents can carry out tasks on their own and use tools, and reasoning models "think" step by step before answering. Today (as of 2026), AI has arrived in almost every professional field, from medicine and law to software development. In 2024, two Nobel Prizes even went to AI research: in physics to John Hopfield and Geoffrey Hinton for the foundations of neural networks, and in chemistry, among others, to Demis Hassabis and John Jumper for predicting protein structures with AlphaFold.

What history teaches us

The history of AI shows a recurring pattern: phases of exaggerated expectations are followed by phases of disillusionment. Today we are in a phase in which AI actually delivers on its promises, but it is important to keep realistic expectations. AI is a powerful tool, not an all-knowing intelligence.

Hype vs. reality: There is still a lot of hype around AI today. Not everything marketed as "AI-powered" lives up to its promise. A healthy dose of critical thinking helps you tell real innovation apart from marketing buzzwords. In the past, exaggerated expectations led to AI winters. Today's AI boom, however, is fundamentally different: the technology already delivers measurable benefits in practice, and development continues to accelerate.

Which event is considered the turning point for deep learning in 2012?

Correct! AlexNet's decisive victory in the 2012 ImageNet competition showed the world that deep neural networks can be used in practice, and it triggered the deep learning revolution.

Not quite. Deep Blue won in 1997, and the Transformer architecture came in 2017. The breakthrough in 2012 was AlexNet's victory in the ImageNet competition, which impressively demonstrated the power of deep learning.

Key Takeaways:

AI research began in the 1950s with Alan Turing's visionary ideas and the Dartmouth Conference.
Two "AI winters" (1974–1980, 1987–1993) resulted from exaggerated expectations and disappointed hopes.
The ImageNet breakthrough in 2012 and the Transformer architecture in 2017 laid the foundation for today's AI revolution.
ChatGPT (2022) made AI accessible to the general public and accelerated development enormously.
History teaches us: critical thinking and realistic expectations are still important today.

This lesson is part of a course with final exam and certificate. View the course "AI Fundamentals"