Ghost in the Machine – A Concise History of AI
These days, we all know what AI (artificial intelligence) is and, most of us even know a little bit about what it’s used for. No longer confined to sci-fi movies, AI is real, it’s here and, it walks among us. In 2019, AI is used in everything from medical procedures to the humble supermarket, helping us to change user experience (or UX) beyond recognition. Although you may know all this (smarty pants!), you may or may not have wondered just where and how this life-changing technology began – but we have. The journey of AI from its conception to the present day has been a long and strange one. Many people mistakenly assume that artificial intelligence, or machine learning, is a modern-day invention and, so, you may be surprised to learn that it began to spread its roots back in the 1950s. The following is our concise history of this vital and ever-changing technology.
In this year, almost 67 years ago, Arthur Samuel wrote a program for playing checkers (draughts) for IBM, which, in 1955, became the first machine learning system to be recognised by the public. Arthur’s invention was showcased on television in 1956 and, it’s said that he coined the phrase ‘machine learning’ in 1959. Arthur continued to work within the realm of AI throughout his career and retained an active interest in machine learning until his death in 1990.
Back when we were all watching Friends and listening to Nirvana, machine learning was going through an incredibly important phase in its development. From 1990 onwards, there was a perceptible shift from knowledge driven machine learning to data driven. This meant that, rather than relying on the logic or knowledge of the user, statistical reasoning was now the driving force behind machine learning.
There are those who say that 2006 is the first time that the phrase ‘machine learning’ came into play (Sorry, Arthur!). The term began to be used worldwide in 2006, largely due to the publication of ‘The Discipline of Machine Learning’ by the Mellon School of Computer Science in July of that year. By this time, machine learning itself was well out of its infancy and was entering a stage akin to the aggression and fast development of a teenager.
Just four short years ago, Microsoft introduced ‘Kinect’ – a motion sensor add-on which was able to track 20 different human features at a staggering rate of 30 times per second. As Bill Gates and his pals were showcasing Kinect, AI was well on its way to adulthood and, began to make its presence known throughout the world; slowly creeping into everyday life.
Machine learning today
Since 2015, machine learning has continued evolving at lightning speed with new innovations showing up almost every day, including chatbots and neural learning. We take a look at some of the ways in which machine learning is being used right now:
Analysis of human behaviour
Businesses who used to throw out generic advertising to try to snag customers have long since cottoned on to the importance of understanding the customer and potential customer in order to deliver a more personalised service. Today, machine learning is used to analyse the way that we shop, browse, work and live in order to adjust the user experience accordingly. Machine learning can quickly analyse and interpret behaviour and turn it into actionable results.
Security – and, in particular, cybersecurity, is a major component of every business and every government as cyber-criminals become more sophisticated. Machine learning is vital in this area as it can perform multiple tasks at a super-fast rate to identify and shut down possible risks.
Machine learning is becoming an integral part of the medical industry due to its ability to crunch data, make comparisons and analyse results as well as learning from previous cases. This means that diagnoses can be made quickly without factoring in human error.
UX and Optimisation
These days, machine learning is used widely in troubleshooting of ecommerce issues such as cart abandonment as well as to help optimise web pages – in particular landing pages – in order to ensure customer retention. Creating strong pages with interesting content is vital in keeping customers on the site – and convincing them to click through to the checkout. Machine learning can help to pinpoint the areas or pages on which customers are losing interest.
The different kinds of machine learning
Supervised learning – this is where objects are tagged in order to steer the machine in the right direction.
Unsupervised learning – machines are fed large amounts of data and are left to learn to ‘sort’ this data.
Reinforcement learning – this process involves the machine learning and evolving by analysing the results of previous actions.
Machine learning works, essentially, in three different ways:
Deep learning – the process of sifting through several layers of artificial neural networks to learn and analyse.
Bayesian networks – this method is an examination of the correlation between things which helps the machine to learn and recognise similarities and differences.
Decision tree learning – the analysis of the features of an object to stimulate recognition and learning.
As we move toward 2020, we can expect machine learning to become an even more integral part of our lives – and we look forward to updating this short history in the next few years!
Technology Journalist from London, currently based in Malaga. For 2 years now, I’ve been writing stories about how our internet works – and how it is changing. From artificial intelligence to UX things are happening today at a pace that can seem bewildering. I am the future-processing.com associate.
July 5, 2020