Close
Close

Published

Don’t buy into the AI job apocalypse hype, for the most part

The hype over AI itself has been overtaken by equally exaggerated fears of job losses on an apocalyptic scale. This August has brought a deluge, encouraged perhaps by the misperception that there is nothing much newsworthy during the summer months of the northern hemisphere. Prophets of doom include some who should know better, or rather perhaps recognize that it is worth predicting the worst just in case that does come to pass, remembering the way many foretold the world would end at the start of the year 2000 because of the Millennium bug.

This of course is a very different situation that will unfold over years rather than a day, but the greatest mistake is to assume that AI represents a radical revolution rather than just part of a continual evolution in automation, which does indeed have an ongoing impact on jobs. Latest to weigh in was the Bank of England’s chief economist Andy Haldane, warning that the UK will need a skills revolution to avoid “large swathes” of people becoming “technologically unemployed” as AI makes many jobs obsolete.

Just the next day, we heard senior members of the country’s legal profession opine that AI risked not just jobs but also the primacy of judges to pass sentences, and possibly even juries to reach verdicts. That at least raised some interesting points over how far automation should penetrate the legal systems, and whether laws should be passed to guarantee that humans will always have the ultimate authority over decisions that affect people’s lives, as has already happened in France.

The wider point though is that automation in general does disrupt the world of work seriously as it has done ever since the industrial revolution and arguably right back through civilization to the emergence of agriculture. It has been a two-edged sword, in many cases liberating people from back breaking or tedious work only to replace it with tasks almost as bad, or in some cases even worse.

The overall trajectory may have been upwards but with great disruptions on the way and in that sense the present furor over AI is no different. What is different now is that automation is making a growing impact on many of the so-called professions, which themselves had emerged as a result of earlier innovations, for example mechanisms in agriculture to reduce need for farm labor.

Law, accountancy and finance are all now under attack from algorithms that can automate many of the tasks, and perform them more accurately with almost total consistency. That is not always a good thing, given the scope for incorporating confirmation bias and other sources of systemic or random error, but at least these are challenges that can be overcome with growing awareness of their existence and impact.

In this case, the threat is to jobs that in some cases are tedious but that people are happy to do because they are very well paid. As a result, some economists and people in the technology sector including Bill Gates, have warned that the finance sector and especially banking has had a negative impact on developed economies by sucking out talent for work that does not need it. The argument is that tech companies do need people of higher intelligence but were finding it increasingly hard to attract them given the huge remunerations they could receive in investment banking and financial trading.

At least that has receded slightly since the credit crunch, as those professions have sunk in public esteem, but it is still an issue. The specter of automation through AI, coupled with demonstrations that algorithms often perform better than humans at these supposedly skilled tasks, has at least started some employers in the finance sector to question the need to compete so strongly on pay for the “highest talent” which has stoked wage inflation in that sector.

It has also revealed how many of the tasks are relatively mundane, and this is a wholly positive aspect because there is the potential to liberate people for more fulfilling creative roles that could achieve greater benefits for their companies or wider society. In that sense, the likes of Haldane are right, but their emphasis is too negative.

His warning that there would be a widespread “hollowing out” of the jobs market, leading to rising inequality, social tension and many people struggling to make a living, was followed almost as a footnote by a call for training to take advantage of the new jobs that would become available.

That last point is correct, and the lesson of history is that societies and governments need to act early to mitigate impacts on professions, although this time the required activities are rather different than before. Automation in mining led to decimation of whole communities, some of which took years or decades to recover because there had been no public-led inward investment to develop alternative trades and skills.

In the professions, such ghettoes do not exist to the same extent, or rather if they do it is in places of work such as the City of London or Wall Street, rather than where people live. In many cases remedies will include programs of internal training so that the people affected can continue working for the same employers and deliver benefits to them. This is most apparent in the legal professions, where it is hard to see a negative side to the impact of AI.

So far, AI’s main application there has been helping cope with the huge burden of paperwork and document analysis which bogs down long and complex cases especially, as well as being part of everyday work. This is often monotonous work and has largely resisted automation so far because it involves a base level of intelligence which only now given greater computational power allied to suitable machine learning algorithms can be computerized. Now that the documents are all digitized they are ripe for algorithmic analysis.

Such software improves efficiency and can review documents to flag them as relevant to a case. Then even more helpfully it can find other relevant documents, analyze them and produce statistically relevant summaries. Lawyers then receive only relevant documents coupled with those summaries, with huge reduction in the workload involved, freeing them to consider cases more deeply and advise clients better. That is already happening, and in that sense machine learning is just the next step in the journey of office productivity-  where earlier word processing and spread sheets made just as profound an impact, without the same bleating about jobs.

Another area ripe for automation is financial portfolio management, where advisers make investment decisions on behalf of individuals or firms. Here too, algorithms are outperforming humans because they can match decisions more accurately to investment profiles, as well as take them more quickly. The overall impact will be cost reduction and better returns for investors, with human advisors having to work harder for their fees by helping with planning, more than the underlying buying and selling decisions.

In some cases, the warnings about job losses often ignore the fact that at present there is an acute skills shortage, as can be the case in the technology and especially healthcare sectors. Then the impact can be to increase the scope of human experts rather than replace them, a good example being radiology where a large part of the workload involves poring over scanned images seeking signals of disease or structural damage, such as a fracture.

With such scans now digitized, this task can be performed by algorithms and feedback so far suggests they usually either match or exceed the accuracy of human consultants, while again being much faster. As in the case of legal judgements, human experts remain the final arbiters of decisions over diagnosis and treatment, but with part of the workload automated, each one can serve more patients effectively. Given a shortage of radiologists, there are no jobs under threat in this case.

The more extravagant promises of AI opening up new avenues of expertise denied to humans remain to be proven at this stage, but great benefits can be achieved just by elevating the level of automation. It is true there will be disruption to some jobs, but that is nothing new.

Close