The AI Revolution

The origins of AI lie in the pioneering work of Alan Turing, the famous British mathematician, who in a 1950 paper proposed the idea of a machine that could simulate human intelligence. In this, he devised the famous ‘Turing Test’ as a measure of a successful simulation, in which humans cannot tell whether they are conversing with another human or a machine. The ‘Dartmouth Conference’ at Dartmouth College in the US in 1956 is often considered the official birthplace of AI as an academic field. Here, the term ‘artificial intelligence’ was first used, setting the stage for a research programme that continues today.

In these early years, there was much optimism. Academic researchers predicted that machines capable of human-level intelligence would be a reality within a generation. However, the limitations of early computing power and complexity of natural language understanding frustrated early expectations.

The ‘AI winter’ began in the late 1980s, a period marked by much scepticism about AI’s ultimate potential.

Early AI systems struggled to scale and adapt to real-world complexities, leading to disillusionment and reduced funding.

The 1990s and early 2000s witnessed a resurgence in AI, driven by strong progress in computing power and scientific developments in the statistical field of machine learning.

This era saw the development of algorithms capable of learning from data, shifting the methodological focus of research from mathematical rule-based approaches to more empirical data-driven approaches.

The victory of IBM’s Deep Blue over chess grand master Garry Kasparov in 1997 was a memorable milestone, symbolising the potential of AI.

‍

A turning point in AI history came with the development of deep learning techniques, particularly the deployment of neural networks, which allowed for much richer non-linear predictions than simple linear regression-based approaches. This was epitomised by the success of AlexNet in 2012, a seminal model that dramatically improved the performance of image recognition.

These advancements led to the emerging use of AI in the economy, from internet search to GPS route planning, both led by Google, to recommender systems pioneered in social network feeds such as Facebook and Instagram. The key to these successes was the ability of deep learning systems to learn and predict using huge amounts of data.

The most recent revolution in AI has been the rise of generative AI, enabled by the development of the Transformer model in 2017. This model, introduced in a paper titled “Attention Is All You Need” by researchers at Google, was a breakthrough in handling natural language. It led to the creation of large language models, such as Open AI’s GPT (Generative Pretrained Transformer), with an unprecedented ability to generate informative, coherent and contextually dependent text.

These generative models have already been applied in a variety of fields, from creating realistic images and text to aiding drug discovery. Their ability to learn from huge datasets and generate new content has opened up vast new frontiers, which leads us to the present juncture.