Fears of robots stealing our jobs are misplaced
It was feared ATMs would displace bank tellers. However, deployment of ATMs reduces bank running costs, allowing banks to open more branches
A widely cited 2013 study estimates that 47 per cent of American jobs are at high risk of being computerised soon
Artificial intelligence is the simulation of human intelligence by machines. The term was coined in a 1955 research proposal stating that “good progress could be made in research to get machines to solve the kinds of problems reserved for humans”. This prediction was over-optimistic at the time and AI garnered a reputation for promising more than it could deliver. An exciting breakthrough was eventually made in 2012 and AI is now making rapid progress. However, many people fear that AI will cause massive unemployment. (The Economist published a comprehensive report on AI last June).
AI is already all around us and expanding rapidly. AI is used for detecting credit card fraud, speech recognition (such as Apple’s Siri), translation (Google Translate), legal discovery, photo search, self-driving cars, automated customer service and so on.
The 2012 AI breakthrough was made in an online contest called ImageNet Challenge, a competition that assesses AI researchers’ success in getting computers to recognise and label images automatically. The 2011 winner correctly labelled 25 per cent of the images, but the 2012 winner correctly labelled 85 per cent of the images using a novel technique called “deep learning”. Further progress allowed the 2015 winner to correctly label 96 per cent of the images, surpassing the human ability level of 95 per cent. And, in May 2016, Google’s AI program AlphaGo defeated champion Lee Sedol in the complex game of Go. The game had been thought too difficult for a computer to master but Google’s AI programme learned to play by observing millions of games.
Deep learning is making rapid progress in training artificial neural networks (ANN). ANNs are computer networks of artificial neurons (mathematical functions) arranged in layers, mimicking the human brain organisation of densely connected biological neurons (brain cells). Information supplied to the input layer is passed on to the next layer where the signals are combined, applying different weights to them and passing the results on to the next layer, and so on and on until the output layer is reached on the other side of the ANN. All the artificial neurons in a layer are connected to all neurons in the layers on either side. The deeper the ANN the greater the ability to recognise subtle information in the input data. Training the ANN involves adjusting its internal weights to give the desired response to particular inputs.
A widely cited 2013 study by Carl Frey and Michael Osborne published by the Oxford Martin Programme on Technology and Employment estimates that 47 per cent of American jobs are at high risk of being computerised soon – including most workers in logistics and transport (such as taxi and delivery drivers), office support (such as receptionists), security guards, sales and services (such as cashiers, counter-clerks), legal discovery, accountancy, etc. Automation by AI will relatively spare high skill, non-routine occupations (such as architects and senior managers) on the one hand and unskilled workers (such as cleaners and burger-tossers) on the other hand.
The most routine jobs are most at risk. In the following occupations the figure in brackets is the probability it will become automated, where 1.0 is certain: dentists (0.004), clergy (0.008), editors (0.06), actors (0.37), economists (0.43), machinists (0.56), typists (0.81), retail sales (0.92), accountants/auditors (0.94).
Great fears were expressed at the start of the industrial revolution that machines would displace most workers from their jobs. However, automating a task, allowing it to be done faster and cheaper, increases the demand for human workers to do associated work that has not been automated. This happened in the weaving trade when machines arrived. A single worker could now run several machines. The amount of cloth produced per weaver per hour increased 50 fold, making cloth cheaper, increasing demand for cloth and creating more jobs. The number of weavers quadrupled between 1830 and 1900.
The same thing is happening today. It was feared ATMs would displace bank tellers. However, deployment of ATMs reduces bank running costs, allowing banks to open more branches. Between 1988 and 2004, urban bank branch numbers increased by 43 per cent in the US and total employment increased. ATMs relieve employees of routine work freeing them for tasks machines cannot do, such as customer services and sales. We probably shouldn’t worry much about being replaced by machines.
William Reville is an emeritus professor of biochemistry at UCC