From the Turing Test to the 2024 Nobel Prize: How Artificial Intelligence Has Developed Over the Decades

This is the first article in a series of publications devoted to artificial intelligence. As a starting point, it is worth setting out the key stages in the development of this technology – from Turing and the Dartmouth conference, through waves of progress and stagnation, to generative models and the 2024 Nobel Prize.

The 1950s and the beginnings of artificial intelligence

Although today the debate on artificial intelligence focuses primarily on generative models, its history began much earlier. The very word “robot”, often functionally associated with AI, was popularised by Karel Čapek’s play R.U.R., staged in 1921; the term derives from a Czech word meaning forced labour. Alan Turing is nevertheless regarded as one of the most important precursors in this field. In his 1950 article Computing Machinery and Intelligence, the British mathematician, known inter alia for his work on breaking the Enigma cipher, posed the question “Can machines think?” and introduced the “imitation game”, later known as the Turing Test. That test examines whether a human is able to distinguish a conversation with another human from a conversation with a computer. The term artificial intelligence itself appeared in John McCarthy’s 1955 grant proposal, and the Dartmouth Summer Research Project on Artificial Intelligence, the workshop organised by him in 1956, is widely regarded as the symbolic starting point of AI as a distinct field of research.

The 1960s and the robot psychotherapist

The 1960s constituted an era of intensive development in academic centres. During that period, game-playing programs were created, as were the first pattern-recognition systems, used inter alia in machine programs capable of playing chess autonomously. Pioneering attempts were also made to conduct human-computer dialogue.

Of particular importance was ELIZA, the chatterbot created by Joseph Weizenbaum in 1966, a precursor of today’s chatbots, including ChatGPT. The program did not “understand” language in the modern sense, but instead operated on the basis of keywords and rules for transforming utterances. Even at that time, however, it was already apparent how readily users attributed understanding, intention and empathy to a machine – and why the issue of the human-machine relationship so quickly became one of the important strands in the debate on AI.

The 1970s and 1980s: a period of stagnation

The development of AI was not, however, linear. The following decades brought both progress and periods of diminished interest, now referred to as “AI winters”. Their causes may be seen in the discrepancy between ambitious promises and practical capabilities, as well as in the constraints resulting from the then state of the art, including insufficient computing power. This period was above all a time of development for expert systems assisting in highly specialised fields, an example being MYCIN, a program created to support the diagnosis of bacterial infections.

The 1990s: AI defeats a world chess champion

A new wave of interest emerged in the late 1990s and was closely connected with the growing importance of the private sector. Corporations such as IBM, Apple and Microsoft began to play an increasingly significant role in research. IBM became particularly renowned for Deep Blue, which in 1997 became the first computer system to defeat the reigning world chess champion, Garry Kasparov. According to IBM itself, the machine analysed 200 million chess positions per second and at the same time performed 11.38 billion floating-point operations. By comparison, IBM’s first supercomputer, launched in 1961, was capable of performing fewer than 500 such operations per second.

The present day: the commercialisation of artificial intelligence

The success of Deep Blue not only revived interest in artificial intelligence. The technology ceased to be exclusively the domain of laboratories and began to operate as an element of everyday digital infrastructure. Roomba, an autonomous vacuum cleaner, was introduced to the market in 2002; Microsoft introduced Kinect technology into its consoles, enabling games to be controlled by body movement; and Siri – Apple’s virtual assistant – was made available to users in 2011.

The real acceleration came after 2010, when three factors converged: greater computing power, access to vast datasets from the Internet, and advances in deep learning. Virtual assistants and recommendation systems became part of the everyday digital ecosystem. The real breakthrough, however, came in 2020, because that was the year in which OpenAI presented GPT-3, a large language model using deep learning to generate content that is increasingly difficult to distinguish from text written by humans. Making ChatGPT available to the general public meant that AI entered everyday use – in office work, information analysis, programming and content creation.

The significance of AI for science was symbolically confirmed in 2024, when John Hopfield and Geoffrey Hinton were awarded the Nobel Prize in Physics for discoveries and inventions enabling machine learning using artificial neural networks. At the same time, following the announcement of the award, Hopfield himself spoke of concerns relating to very powerful systems whose operation and controllability we still do not understand sufficiently well. Today, artificial intelligence has not only technological significance, but also legal, organisational and social significance. It affects the labour market, the circulation of information, and decision-making processes. The scale of these issues is already clearly illustrated by specific studies: in the TruthfulQA benchmark, the best-performing large language model tested was truthful in only 58% of its answers, whereas humans achieved 94%, which demonstrates the persistence of the hallucination problem (Lin, Hilton, Evans, 2022); similarly, a study published in Science showed that a widely used healthcare algorithm covering millions of patients systematically understated the health needs of Black patients- after correction, the percentage of persons qualifying for additional care increased from 17.7% to 46.5% (Obermeyer et al., 2019). Its development therefore gives rise to concrete challenges for the future – from ensuring transparency and accountability in the operation of systems, through limiting infringements of copyright and privacy, to the need to build competence and knowledge in relation to AI (so-called “AI literacy”).