Whether it’s film reviewers discussing Blade Runner’s replicants, or computer scientists cheering their algorithm as it beats a human at yet another game, everyone’s talking about artificial intelligence. The government certainly has recognised the incredible opportunities presented. A review by professor Dame Wendy Hall and Jérôme Pesenti has estimated that AI could add an additional £630 billion to the UK economy by 2035.
The term AI encompasses algorithms that, without being explicitly taught how to do so, learn how to perform a particular task, optimising their performance as they go. The mathematics underpinning these algorithms is not new, but the explosion of data and advances in computing power in the last few years means that AI algorithms can now undertake a greater range of tasks than ever before, including those ‘cognitive’ tasks that were previously thought to be the preserve of humans. From filtering email spam, to delivering takeaways, to providing legal advice, the speed at which AI is automating a wider range of tasks would appear to justify all the recent hype.
The best and most recent estimates of the automatability of different employment sectors come from PwC. Unsurprisingly, the risks appear highest in sectors such as transportation, storage, manufacturing and retail. What we have done is apply these risks of automatability to the sectoral make-up of each British parliamentary constituency, to better understand how specific geographical locations could fare as automation marches on.
Our findings are striking. Although PwC put the percentage of jobs at high risk of automation at 30 per cent for the whole of the UK by the early 2030s, we have found that these proportions vary significantly across Great Britain, from 22 per cent to over 39 per cent. Former industrial heartlands including in the Midlands and north of England will be hit hardest by job losses. In fact, of the top 50 constituencies with the highest proportions of high risk jobs, over two-thirds are in these regions. At individual constituency level, shadow chancellor John McDonnell’s constituency of Hayes and Harlington has the highest percentage of jobs at high risk of automation, at almost 40 per cent. What the top 5 constituencies have in common is a reliance on jobs in transport, storage and manufacturing (Hayes and Harlington includes Heathrow Airport, for example).
The speed of job displacement is likely to dramatically outstrip that of previous technological revolutions. In the third most impacted constituency, North Warwickshire, 20,500 jobs could be displaced by AI-driven automation by the early 2030s. Compare these losses over the next 15 years with the several decades it took for the disappearance of the 19,000 mining jobs in the whole of Warwickshire at the industry’s peak in 1913. It is concerning that the areas that have already suffered so much from industrial decline could be hardest hit yet again.
And this is without even considering the other potential inequalities in the impact of automation. Indeed, for individual workers, the key differentiating factor in determining their risk of job displacement is education. For those with GCSE-level education or lower, the estimated potential risk of automation is as high as 46 per cent, falling to only around 12 per cent for those with degrees. Similarly, men may be at higher risk of job displacement by automation than women.
Yet, in spite of all this potential disruption, our latest YouGov online poll reveals that only 7 per cent of UK adults are worried that their current job role will be replaced by AI, and only 28 per cent are worried about jobs in their local area.
What should the government be doing? Firstly, a ‘one-size-fits-all’ approach simply will not work. We need more research to better identify those who will be worse affected, and targeted interventions that ensure these impacts are minimised, such as providing financial and psychological support. The education system should be reformed to focus more on automation-resistant abilities such as creativity and interpersonal skills, while jobs where these skills are critical, such as nursing and social work, should be made more attractive and better rewarded. Serious thought should also be given to studying alternative taxation and income models that result in fairer distribution of the wealth these technologies will generate.
Technology has undoubtedly changed the world for the better in many ways, and artificial intelligence is no different in its potential to improve everyone’s quality of life. History will judge these technologies as not having fulfilled their promise, however, if they worsen the inequality that blights modern society. We need action now to make sure that automation works for everyone.