A brief history of research in Artificial Intelligence (AI) began in 1956 at Dartmouth College Summer School in the US, where it was thought that 10 scientists could, in two months, lay the foundations for a machine capable of using language, handling abstract concepts, solving problems that had previously been the exclusive domain of humans, and improving itself. Today, their optimism may be laughable, and it is not surprising that they have made much slower progress than planned, but the point is that they have taken the first, crucial steps.
Winters and Summers
The peaks and troughs of AI research are known as “winters and summers”. The hype and promising results – ‘summers’ – are usually followed by failures and criticisms, which go hand in hand with the drying up of funding sources and longer periods of unproductive periods – ‘winters’. The summer we are enjoying now was ushered in by the advent of the internet, which enabled the explosion of data known as ‘big data’.
The big breakthrough
The realisation that intelligence is not the same as knowledge was a turning point in the research into artificial intelligence, because so many things humans do by instinct rather than by rule (e.g. we don’t do physical calculations when we catch a glass falling from a table) and it would be impossible to program all the knowledge and rules into machines anyway. Intelligence is more about learning and the ability to generalise from learning.
However, effective learning requires a sufficient amount of data, and around 2012 we reached a point where the combination of machine learning and big data gave rise to what is known as deep learning. This has led to a number of spectacular results in our daily lives, such as the dramatic improvement in the accuracy of voice recognition systems, increasingly reliable self-driving cars, highly efficient facial recognition systems (which not only act as the ‘Big Brother’ of totalitarian states, but also find wanted criminals and lost children), or the fact that humans can now compete for second place in chess and go.
The importance of data
Thanks to deep learning, artificial intelligence can now be used to solve an extremely wide range of problems. However, learning models requires a lot of high-quality data, which is often harder to acquire than the AI technology itself and the computing power to build and run AI models. AI is therefore no longer the future, but the tool of the present. For building AI solutions, feel free to get contact to the experienced and enthusiastic team at Neuron Solutions.
