A brief history of AI

Early artificial intelligence: 1950s – 1960s

Artificial intelligence, or AI, has been a subject which has fascinated people since the first days of digital computing. There have been countless books, films and stories written about artificial intelligence, and in the 1950s and 60s, there was a particularly huge interest in how artificial intelligence would change our world!

The Highgate Wood telephone exchange, From Wikimedia, CC-BY-SA 3.0

In the early days of AI in the 1950s and 1960s, computers were enormous machines, the size of a whole room. They were controlled by manually changing the hardware, which back then included a lot of wires and vacuum tubes;  advanced machines could be programmed using punch cards.

Computers were very uncommon; they were used by universities for research, by the military, and by governments to improve infrastructure such as the Highgate Wood telephone exchange machine which is pictured above. This was the first automated telephone exchange. It was able to connect telephone calls, a job which had previously been done by human operators.

In those early days, computers were developed which could play chess and solve maths problems. It was hoped that artificial intelligence would be commonplace in people’s homes by the 1970s, with robot housekeepers and workers which would help to rebuild the workforce after the Second World War.  As you will know, however, this was not to be, and we are still waiting for many of the promises made by those early artificial intelligence pioneers to be realised.

The first AI winter: 1970s

During the 1970s, computers were slowly growing more powerful and increasingly common, and were starting to become more similar to the computers which you may be familiar with today. However, the progress of artificial intelligence was very slow and both researchers and the general public had more or less lost interest in it. The utopian future of home assistants and robot workers which people had dreamed of in the 1950s and 1960s had not appeared, and the funding available for researchers of artificial intelligence decreased.

Expert systems: 1980s

By the 1980s, computers had grown far more advanced, and were now commonly found in homes, schools and workplaces. The internet wouldn’t emerge until the 1990s and wouldn’t become common in peoples’ homes until late in the decade, but computers were now an everyday item in many countries.

Researchers tried a new approach to AI, which was to make ‘expert systems’. These programs were designed to ‘think’ like experts working in fields such as finance and medicine, and to make decisions in the same ways as those experts. This was a huge success, and companies which made expert systems realised that these were very in-demand!

The finance industry was being revolutionised by the systems, which were able to automatically detect fraud and other crimes. These systems were also able to predict stock market changes, and soon many investment companies were using expert systems to manage huge amounts of money, as they were able to invest more profitably than humans could.

Machine learning and neural networks: 1990s

The success of expert systems gave AI a new boost of interest and funding. AI systems were now learning from data, which is a field of AI called ‘machine learning’. Programmers were making neural networks which are a technology inspired by how brains work, and these could learn from data, recognising patterns and using those patterns to make predictions.

This began to have a significant impact in lots of industries. Artificial intelligence models were now often monitoring other processes such as manufacturing, able to predict when machines would need maintenance, for example.  Factories were becoming much more automated too, with sensors and powerful computers able to work with very little human interference. This was great for manufacturers, but also began to have negative impacts on workers with many factories reducing their staff numbers.

The internet and Big Data: 2000s

At the start of the new millennium, technology was booming. The internet was becoming accessible to larger parts of the population in economically developed countries, and people were creating huge amounts of data as they browsed, shared and interacted online. As smartphones emerged in the late 2000’s, people were generating more and more data online, and were also interacting with artificial intelligence more and more.

Artificial intelligence was being used in lots of areas online; advertisers were using data to target their adverts, search engines were improving their results, and researchers were using artificial intelligence and machine learning to find patterns in this data. AI was being used in every industry, from education to baking.

Deep learning and AI assistants: 2010s to the present

With this new wave of interest, AI tools were becoming increasingly common in peoples’ lives. With AI helping to make speech recognition easier for computers, tools such as Siri, Google Assistant and Amazon’s Alexa were now able to recognise different voices and to understand their users better than ever.

Big breakthroughs in image classification meant that driverless cars were becoming a possibility. Image classification is the process of an AI model looking at an image and working out what it contains, which is surprisingly difficult for machines! With these new developments, companies such as Tesla began developing automated vehicles. These were able to understand signs, road marking, and recognise dangers such as pedestrians and other vehicles, and to respond to them like a driver would.

Large language models (LLMs) such as ChatGPT began to appear. These tools learn from huge bodies of text and can understand and respond to prompts from the user, allowing them to give helpful (and sometimes less helpful!) advice. Along with the large language models, image models were also able to produce images from text prompts.

The future

AI is currently experiencing a huge boom of popularity. It is now much more accessible to the wider public, meaning that it can be used in a much wider range of fields and jobs than ever before. As AI continues to progress, it could help us to solve some of the big problems affecting our world too!

Next steps

Having looked at a brief history of AI, we will now move on to look at one of these developments in particular, the one that is arguably having the most impact currently: large language models.