The history of artificial intelligence (AI) from ancient times to the present

Author: Jozef Wagner

Digitization

12. 07. 2024 20 reading minutes

Artificial intelligence (AI) is currently a hot trend and is becoming an everyday part of our lives. It brings technological advances, fresh wind to many industries and changes the way we learn, live or work. That this is no short-term fad is evidenced by the fact that NVIDIA, a company that produces AI chips in bulk, recently became the world’s most valuable company. Leading IT firms are investing huge sums of money in AI infrastructure in order to have the opportunity to participate in the research, development and training of advanced AI.

However, this technology will slowly make its way into the homes of ordinary people, as leading processor maker AMD will incorporate a Neural Processing Unit (NPU), a kind of co-processor for AI computing, into its new processors. What it will be used for is still hotly debated.

AI technology has long since beaten humans at chess, quizzes or the oldest board game , GO. Today, it literally leaves people amazed at how it can render beautiful video from descriptive text, create breathtakingly original scenery, compose music in the style of artists who are no longer living, translate in real time between two foreign languages, tutor children in math, and much, much more. Some people see AI as a technological marvel, and would probably be properly surprised to know that artificial intelligence isn’t actually that new. It’s been with us for decades. The mathematical models and theory on which it operates have waited until today to be allowed to shine in its best light, in the form of supercomputers.

In the last article, we introduced the concept of artificial intelligence. You can read it here – What is Artificial Intelligence? Today, we’ll take a look at when and how AI came to be, introducing the most important milestones that have shaped it into what it is today.

From ancient times to the Middle Ages

Already in antiquity (a few thousand years ago) ancient philosophers debated the questions of life and death, and it was in this period that inventors began to create mechanical toys, which they called automata (automatons). Automaton, hereafter referred to as automaton, comes from ancient Greek and means “acting of one’s own will”. It was a simple mechanical thing that moved without human intervention. As time went on, more and more advanced mechanical machines began to emerge, the most famous of which I would mention being Da Vinci’s knight. From this time period comes the idea of a machine functioning on its own.

17th to 19th century

In the 17th century, the philosopher Rene Descartes began to theorise that one day machines would be able to think and make decisions. In 1637, he wrote down his ideas in a book Discourse on Method. He divided machines into those that could one day learn to perform one specific task and those that could adapt to any job. Today, these areas are known as specialized and general AI. In a way, this introduced the challenge of creating AI.

This period was also rich in mathematical discoveries.

1642: Blaise Pascal invents the first mechanical calculator. This could add and subtract two numbers, multiplication and division worked by repeated addition or subtraction.

1676: Gottfried Wilhelm Leibniz derives the chain rule. This rule is used by AI to train neural networks with backpropagation.

1738: Daniel Bernoulli introduces the concept of a utility function. It is a generalization of probability and the mathematical basis by which intelligent agents represent their goals.

1739: David Hume described induction, a logical method for learning generally valid propositions from examples.

1763: Thomas Bayes lays the foundation for Bayes’ theorem, which is used in modern AI for Bayesian networks.

1837: Charles Babbage and Ada Lovelace create the first design for a programmable machine.

1854: George Boole invents the famous Boolean algebra.

1859: Charles Babbage and Ada Lovelace worked on programmable mechanical calculating machines focused on polynomial functions.

1863: Samuel Butler came up with the theory that Darwin’s evolution also applies to machines, and perhaps one day they will gain consciousness and eventually replace mankind.

First half of the 20th century

In the early twentieth century, science fiction authors and scientists began to wonder if it was possible to create an artificial brain. Some inventors began creating humanoid characters that were, for the most part, powered by steam and some could walk. Experimentation with facial expressions also began.

1921: Czech writer Karel Čapek was the first in the world to use the word robot in his science fiction play RUR. He used it to refer to an artificial human.

1929: Japanese professor Makoto Nishimura builds the first Japanese robot, calling it Gakutensoku.

1943: Warren Sturgis McCulloch and Walter Pitts published the first mathematical description of an artificial neural network in a scientific paper titled A Logical Calculus of the Ideas Immanent in Nervous Activity.

1944: Game theory emerges and becomes an important part of AI development.

Before 1949, large hall calculating machines functioned as calculators. They could not store commands, only execute them. Plus, pretty expensive calculators, monthly rentals climbed to $200K a month. Only prestigious universities and large technology companies could afford them. Many specific calculations, such as the trajectory of a rocket launch, were calculated by teams of mathematicians on paper.

Alan Turing

Alan Turing was a pioneering British mathematician and computer scientist who is often regarded as the father of artificial intelligence. In 1950, Turing devised a test, later known as the Turing Test, to determine a machine’s ability to exhibit intelligent behaviour indistinguishable from human behaviour. In this test, a human judge engages in a conversation with a human and a machine, both of which are hidden from view. If the judge cannot reliably distinguish who is who, the machine is deemed successful and demonstrates human-like intelligence. If the machine fools the human into thinking it is a human, then it is intelligent.

1951: Marvin Minsky and Dean Edmonds develop the first artificial neural network (ANN) called SNARC using 3000 vacuum tubes to simulate a network with 40 neurons.

1952: Computer scientist Arthur Samuel created a program to play Checkers, which he perfected in 1955 so that the program taught itself to play.

Dartmouth Conference AI (1956)

The Dartmouth Conference, held at Dartmouth College in the summer of 1956, is considered the seminal event that defined artificial intelligence as a new field of study. Organized by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon, the conference brought together leading researchers to discuss and explore the possibility of creating intelligent machines. It was at this conference that the term ‘artificial intelligence’ was coined and laid the foundations for future research and development in AI, setting out the key goals and concepts that will shape the field in the decades to come. Many of the areas that underpin AI today, including natural language processing, computer vision and neural networks, were also part of the agenda.

Perceptron AI (1958)

Perceptron, developed by Frank Rosenblatt in 1957, is one of the first artificial neural network models and a fundamental milestone in AI. It was designed to simulate the thought processes of the human brain and could learn to classify input data into different categories. Perceptron demonstrated the potential of machine learning and neural networks, laying the foundation for modern deep learning techniques. The neural network contained only one layer, and despite its limited functionality, these concepts were the cornerstone for subsequent neural network research.

1958: John McCarthy develops the Lisp programming language, which has become hugely popular among AI developers.

1959: John McCarthy and Marvin Minsky found the Artificial Intelligence Research Institute MIT AI Lab.

In this year, Arthur Samuel coined the term machine learning when he mentioned in a speech machines that play chess better than the humans who programmed them.

1961: General Motors deploys the first industrial robot , Unimate, to replace humans on the assembly line.

ELIZA AI (1966)

ELIZA was the first chatbot to simulate human conversation. It was created by Joseph Weizenbaum. Although the capabilities of this chatbot would not appeal to anyone today, ELIZA demonstrated the potential for natural language processing in AI. This computer program simulated conversation using pattern matching and pattern replacement to give users the illusion of understanding. ELIZA demonstrated the potential for machines to engage in dialogue and that computers can generate human-like responses. This chatbot sparked interest in the field of conversational AI.

1966: Shakey is the first universal mobile robot that reasons its own actions. It was equipped with sensors and a TV camera, which it used to navigate through different environments.

1967: Newell and Simon develop the General Problem Solver (GPS), one of the first artificial intelligence programs to demonstrate human-like problem solving.

1974: The first AI winter begins, marked by a decline in funding and interest in AI research due to unrealistic expectations and limited progress. AI has so far been viewed as a nice tech demo with limited real-world utility.

1979: The American Association for Artificial Intelligence is founded, now known as the Association for the Advancement of Artificial Intelligence.

In the same year, the first computer-controlled autonomous vehicle , the Stanford Cart, passes a chair obstacle test.

In the 1980s, there was a renewed interest in AI and investing in it. Deep learning techniques and the use of expert systems became popular; these allowed computers to learn from their mistakes and make independent decisions. Companies began to make huge savings from expert systems, and by 1985 corporations were investing a billion dollars a year in AI systems.

1980: The first expert system, known as XCON, came to the commercial market. It was designed to assist in the ordering of computer systems by automatically selecting components according to the customer’s needs.

Other expert systems are following suit, gaining popularity as companies use them mainly for financial forecasts and medical diagnoses.

1981: Danny Hillis designed parallel computers for AI and other computational tasks, an architecture similar to modern GPUs.

1986: Hinton, Rumelhart, and Williams publish Learning Representations by Back-Propagating Errors, which allows more complex neural networks to be trained.

In the same year, Ernst Dickmanns, a scientist working in Germany, invented the first autonomous car. Technically a Mercedes van that was equipped with a computer system and sensors to read its surroundings, the vehicle could only drive on roads without other cars and passengers.

By the late 1980s, interest in AI was waning, and another winter came to AI because of high costs versus seemingly low returns. This term describes a period when research funding is reduced due to low investor and consumer interest. But this is about to change.

1991: It is the year of the birth of the Internet. CERN researcher Tim Berners-Lee launches the world’s first online website and publishes the Hypertext Transfer Protocol (HTTP). Although many institutions and large enterprises had already networked computers before, the advent of the Internet became a huge boost to society, and within a short period of time, millions of people from different parts of the world connected to the Internet and began generating the quanta of data that would later become a much-needed commodity for AI training.

Deep Blue vs Kasparov (1997)

Deep Blue, developed by IBM, made history in 1997 when it was the first computer to defeat reigning world chess champion Garry Kasparov in a six-game match. Deep Blue’s victory marked a significant milestone in the field of AI and demonstrated the power of brute force computing and advanced algorithms in solving complex problems. The match demonstrated the potential of AI to outperform human experts in specific domains, sparking tremendous interest in advancing AI research and development. You can find the Deepblue vs Kasparov video here.

1998: Cynthia Breazeal at MIT introduces KiSmet, an emotionally intelligent robot that recognizes and can respond to people’s feelings.

1999: Sony launches AiBO, the first pet dog whose personality and abilities develop over time.

2002: iRobot introduces Roomba, the first mass-produced home robotic vacuum cleaner with an AI-powered navigation system. It moves autonomously and avoids obstacles when vacuuming.

NASA sends rovers to Mars (2004)

Mars orbited much closer to Earth in 2004, so NASA took the opportunity to send two rovers – called Spirit and Opportunity – to the red planet. Both were equipped with AI to help them navigate Mars’ difficult rocky terrain and make decisions in real time, instead of relying on human help.

NASA’s rovers on Mars, such as Spirit, Opportunity and Curiosity, represent significant milestones in AI and robotics. These rovers are equipped with sophisticated autonomous navigation systems that allow them to explore the Martian surface, conduct science experiments, and make independent decisions to avoid obstacles and identify points of interest. Space missions launched between 2004 and 2012 have greatly improved our understanding of Mars while demonstrating practical applications of AI in autonomous systems, remote control, and scientific discovery in extraterrestrial environments. Click here for a video of the Spirit rover landing on Mars.

2006: Twitter, Facebook and Netflix start using AI as part of their advertising and user experience algorithms.

2009: Google builds the first self-driving car that can handle city driving.

Watson the computer wins the quiz show Jeopardy! (2011)

In 2011, IBM’s Watson made headlines when it won the quiz show Jeopardy! against two of its biggest champions, Ken Jennings and Brad Rutter. Watson’s victory demonstrated AI’s advanced capabilities in natural language processing, information retrieval, and machine learning. The system was able to understand and respond to complex questions posed in natural language, quickly analyse vast amounts of data and generate accurate answers. This achievement demonstrated the potential of AI in processing and understanding human language, marking a significant milestone in the development of AI applications and technologies. You can find a video of Watson on Jeopardy here.

2011: Apple integrates Siri, an intelligent virtual assistant with a voice interface, into iPhone 4S.

After 2011, AI Deep Learning and Big Data techniques are coming to the fore. Significant advances were made in image recognition in 2011 and 2012, mainly due to new hardware that increased the learning speed by up to a hundred times. At that time, machine learning was done on graphics chips, which were ideal for this because they optimized the handling of vectors and matrices.

2012: AI startup DeepMind develops a deep neural network that can recognize cats in YouTube videos. Google later bought DeepMind for $500 million in 2014.

That same year, Facebook creates DeepFace, a facial recognition system that can recognize faces with 97% probability – an accuracy close to humans.

Generative adversarial networks (2014)

In 2014, Ian Goodfellow introduced Generative Adversarial Networks (GANs), which revolutionized the field of AI. GANs consist of two neural networks, a generator and a discriminator, which are simultaneously trained through a competitive process. The generator produces synthetic data, while the discriminator evaluates its authenticity against real data. This adversarial process enhances the generator’s ability to produce highly realistic outputs, leading to advances in image and video generation, data augmentation, and many other applications. GANs have had a significant impact on AI research and have opened new avenues for creativity and innovation in machine learning.

2014: Amazon launches Alexa – an intelligent virtual assistant with a voice interface that completes shopping tasks.

At the same time, chatbot Eugene Goostman passes the Turing test, with a third of the judges believing Eugene is human.

2015: Computers can identify objects in visual data much more accurately than humans can. In the annual ImageNet challenge, they achieved an accuracy of 97.3%, which compares to only 71.8% in 2010.

AlphaGo vs Lee Sedol in Go (2016)

AlphaGo, developed by DeepMind, made history in 2016 by defeating Lee Sedol, one of the world’s best Go players, in a five-game tournament. It was a groundbreaking achievement because Go is an incredibly complex board game with over 100 thousand opening moves, of all the possible positions there are even 2 per 170, making brute force methods unusable. AlphaGo’s success has been achieved by its advanced machine learning techniques, including deep neural networks and reinforcement learning. This milestone demonstrated AI’s potential to handle highly complex tasks, significantly advanced the field of artificial intelligence, and demonstrated its ability to solve problems previously thought to be beyond the reach of machines.

2016: Hanson Robotics created a humanoid robot named Sophia, which was the first robot created with a realistic human appearance and the ability to see, joke, communicate as well as replicate emotions. Thanks to her innovative AI and abilities, Sophia became a global phenomenon and regularly appeared on talk shows, including late night programs like The Tonight Show.

2017: Google’s AlphaStar beats the best dedicated chess engines in a series of matches.

2018: Artificial Intelligence is taught in most universities.

2019: AlphaStar has reached Grandmaster level in the computer game StarCraft 2, meaning that with its AI it can defeat a respectable 99.8% of human players.

The rise of generative AI (2020 – present)

Developments in the field of generative AI have caused a surge of interest in AI in recent years. Generative AI offers the ability to generate text, images and videos in response to text prompts. Unlike previous systems, which were programmed to respond to a set prompt, generative AI learns from materials (documents, photos, and more) and data from across the internet.

Language model GPT-3.5 (2020)

GPT-3, developed by OpenAI and released in 2020, is one of the largest and most advanced language models ever created, with 175 billion parameters. It can generate human text based on given prompts, perform a variety of linguistic tasks such as translation, summarization and question answering, and even produce coherent and contextually relevant content. GPT-3’s ability to understand and generate natural language has set a new standard in AI, demonstrating the power of large-scale neural networks and transforming the way machines interact with human language in a variety of applications.

2021: OpenAI has developed DALL-E, which can generate digital images directly from text using deep learning methodologies.

At the same time, DeepMind’s AlphaFold2 solves the problem of protein folding, paving the way for new drug discovery and medical breakthroughs.

ChatGPT OpenAI (2022)

ChatGPT, developed by OpenAI and launched in November 2022, is a breakthrough conversational AI based on the GPT-3.5 architecture. It excels at generating coherent and contextually relevant text that enables natural and dynamic human-computer interactions. ChatGPT can perform a wide range of tasks including answering questions, providing detailed explanations, and assisting with creative writing. Its advanced capabilities have revolutionized applications in customer service, education, and content creation, demonstrating the transformative potential of AI in improving everyday human-computer interactions.

January 2023: ChatGPT became the fastest growing app, with over 100 million users.

March 2023: OpenAI releases a new version of the GPT-4 model, which now allows you to enter an image as input instead of text. Google introduces its Gemini chatbot.

February 2024: OpenAI publicly introduces Sora, which offers the ability to generate videos from text up to one minute in length.

May 2024: GPT-4o AI technology was introduced, doubling the speed of the API, breaking all benchmark records, and finally bringing emotion to the chatbot conversation.

AI and the future

We’ve been through the fascinating historical development of AI. What does the future hold? We’ll see. All that is certain is that modern AI technologies will begin to make a big push in both businesses and homes. Many jobs will disappear or will have to adapt to the new AI trend. But on the other hand, many new jobs will be created. Advances in AI are unstoppable and it is up to us to harness them for the common benefit of humanity.

Share article

The history of artificial intelligence (AI) from ancient times to the present

Jozef Wagner

I have been programming in Java for more than 10 years, currently I work at msg life Slovakia as a Java programmer senior and I help customers implement their requirements into the Life Factory insurance software. In my free time I like to relax in the forest or play a good computer game.