Artificial intelligence: its origins and evolution

Have you ever wondered how the intelligent machines that surround us came about? Artificial intelligence, a term that summons images of humanoid robots and complex systems, has roots that reach back to some of the brightest minds in history.

Where it started

The idea of building machines that can think and learn is not new. Philosophers and mathematicians in ancient Greece already speculated about intelligent automata. It was not until the middle of the twentieth century, though, that artificial intelligence took shape as a formal field of study.

• Alan Turing and the Turing test: in 1950 the British mathematician proposed a test to determine whether a machine could show behaviour indistinguishable from a human. That test laid the ground for conversational systems and natural language processing.
• The summer of 56: in 1956 a group of scientists met at Dartmouth College to discuss the possibility of building machines that could think. That meeting marks the official birth of artificial intelligence as a research field.

The early years and what stopped them

The first years were marked by large ambitions and real progress. Researchers built the first programs capable of playing chess and proving mathematical theorems. They also ran into serious obstacles:

• The AI winter: through the nineteen seventies and eighties, funding for AI research fell sharply for lack of tangible results.
• Computational limits: the computers of the period were not powerful enough to run the algorithms the field needed.

What brought it back

• Computing power: decades of steady gains produced machines that were fast enough and cheap enough.
• Availability of data: the internet generated the volume of data that training requires.
• New algorithms: deep learning changed what the field could attempt.

The kinds of artificial intelligence

• Narrow artificial intelligence: the common case today. It is built for specific tasks such as facial recognition, machine translation or autonomous driving.
• General artificial intelligence: a machine able to understand, learn and apply knowledge to any intellectual task a human can perform. We remain far from that.
• Superintelligence: an intelligence that would surpass the human one in every respect. The concept is the subject of considerable speculation and debate.

Machine learning and deep learning

• Machine learning: lets machines learn from data without being programmed explicitly. The algorithms identify patterns in large datasets and use them to predict or to decide.
• Deep learning: a subset that uses artificial neural networks to model complex relationships in data. Inspired by how the brain works, it has produced remarkable results in image recognition and language processing.

Generative AI

Generative AI creates new content: images, music, text or code. It uses generative models and neural networks to produce data that is hard to distinguish from the real thing. Some of its applications:

• Image generation: realistic images produced from a textual description.
• Text generation: creative writing, from verse to scripts.
• Music generation: original composition.

Where the field is going

• Democratisation: tools and platforms grow more accessible, which lets a wider public build and use AI applications.
• AI everywhere: it is being integrated into every part of daily life, from telephones to health systems.
• Ethics: as AI grows more powerful, the questions about how it is built and used stop being academic.

Artificial intelligence is a field in constant movement with enormous potential to change how organisations work. Machine learning, deep learning and generative AI are only some of the technologies driving it.