IP24: AI, Skills and Jobs. Five Reasons for Optimism
Many thanks for the insightful comments on IP23 regarding the politics of techno-optimism. Clearly, this is an important issue that sparks a great deal of interest. It's a topic I'll be revisiting in future newsletters, but please continue sharing thoughts. This recent article in Politico about “regulation fatigue” in Brussels suggests that even the arch regulators might be looking for a corner to be turned from techlash.
One of the hallmarks of techno-optimism is the well-grounded belief that AI will increase productivity. Conversely, AI-doomerism is characterised by the fear of AI replacing jobs, especially white-collar positions.
This week, I'd like to explore the emerging idea that the optimists are doubly right on this front, and the doomsters might be doubly wrong. In short, the argument goes that AI will not only increase productivity but also create new jobs, with the crucial aspect being that AI will facilitate wider access to the skills needed to thrive in these new positions. Additionally, AI could revitalise sectors like manufacturing, which hold a broader emotional connection with the public and offer jobs that provide a sense of purpose and fulfillment.
AI, Skills and Jobs. Five Reasons for Optimism
In particular, there are five reasons for optimism about AI and jobs, which I’ll consider in turn:
AI is already producing productivity gains. This will lead to increased growth and more jobs. These gains have come when AI has been complementary to existing work, rather than replacing it.
Like previous waves of automation, AI will create new types of jobs.
Big productivity gains are likely to be in strategically important sectors that provide benefits across the economy, such as manufacturing and agriculture.
AI might go some way towards tackling the hourglass economy, dominated by knowledge jobs at the top and insecure jobs at the bottom.
Unlike in previous waves of automation, governments and industry are serious about tackling skills needs.
AI is already producing productivity gains. Further productivity gains will boost growth and jobs.
There are a number of estimates about the boost that AI can bring to already developed economies. Goldman Sachs, for example, suggests that AI could raise global GDP by up to 7 per cent. McKinsey has argued that the impact of AI could be to increase global productivity by up to 3.3% a year, when combined with other elements of automation. Some observers even suggest that AI could contribute to “explosive growth” of the kind seen during the Industrial Revolution.
Brookings have argued that we could see an AI powered productivity boom, arguing that productivity growth would compound as innovation continues and there is further acceleration in the kind of R&D that delivers greater gains in productivity. The nature of AI and machine learning means that scientific discovery will also accelerate over time. The figure below is taken from the Brookings document.
Although the scale of the boost is disputed, it’s widely agreed that increasing integration of AI will boost both productivity and GDP growth. Productivity growth has historically allowed for an increase in economic output, along with higher wages and improved living standards. As the World Bank has suggested, “productivity growth has led to overall jobs growth, as increased demand for workers has outweighed technology’s replacement effects.” Economist, Austin Goolbee, puts it vividly when he says that productivity represents, ““magic beanstalk beans for the economy. … You can have faster income increases, faster wage growth, faster GDP without generating inflation.’’
And there are signs that AI is already delivering some productivity gains, even before wide-scale implementation, with a Bloomberg report suggesting that AI investments will add 0.5% to US GDP this year and Fortune saying that AI is already delivering a “boom in productivity”, which has allowed the United States to avoid the much predicted recession and increase in unemployment. Crucially these productivity gains have been shown to be additive or complementary, rather than replacing, with a National Bureau of Economic Research study finding that “the [productivity boost] effect was largest for the least skilled and least experienced workers, who saw productivity gains of up to 35%.” This is partially due to AI’s ability to utilise the knowledge of the most-skilled workers and disseminate across the wider workforce.
This goes to the heart of the productivity and growth malaise that has impacted large parts of the West, and particularly Europe, for the past few decades. The UK, for example, has seen its productivity lag behind comparable countries. Since 2000, the Eurozone’s growth has lagged behind the United States by 17 per cent.
Europe and many parts of the West face a public policy conundrum, where growth is stagnating just at the time that demands of an aging population mean that pressures on public services and the public purse continue to grow. The productivity boost that is likely to come from AI is, therefore, not only crucial to boosting economies, but also essential to maintain the European social model over the medium-term.
Like previous waves of automation, AI will create new types of jobs.
The type of work available in an economy does not remain static as technology advances. Instead, new technology brings with it new types of work. The Industrial Revolution, driven by disruptors like George Stephenson (profiled in IP 14) led to jobs that would previously have been unthinkable. The computer, the internet and the smartphone all created new jobs and new categories of jobs. And there seems little doubt that AI will do the same. Indeed, in the second quarter of last year, the number of AI related job postings on a major job board increased by over 1,000%.
Ben Evans considered the argument in his excellent newsletter. He suggested that:
Every time we go through a wave of automation, whole classes of jobs go away, but new classes of jobs get created… we know (or should know), empirically, that there always have been those new jobs in the past, and that they weren’t predictable either: no-one in 1800 would have predicted that in 1900 a million Americans would work on ‘railways’ and no-one in 1900 would have predicted ‘video post-production’ or ‘software engineer’ as employment categories…
He draws upon the Lump of Labour fallacy to remind us that there is no “fixed amount of work to be done, and [the fallacy] that if some work is taken by a machine then there will be less work for people.” Instead, automation has historically increased productivity in the economy, boosted efficiency, spread prosperity and created new jobs
Evans and others also call upon the Jevons paradox from the 19th Century, devised by British economist William Jevons in 1865, which dissected traditionally held ideas that increased efficiency would reduce use of coal, in particular. In his book, The Coal Question, Jevons pointed out that as steam engines became more efficient (and used less coal), steam engines would be used in a greater range of industries, leading to greater demand for coal and a growing economy. He was, of course, right and since then innovation has created more jobs across industries. There is little reason to think that AI will be different.
The World Economic Forum predict that AI will create 97 million new jobs by 2025, which is 12 million more than the number of jobs that it expects AI to displace by 2025. They produced the fascinating chart below setting out those areas of work likely to grow and those more exposed to displacement in the coming years:
It’s also now pretty clear that AI is likely to be additive, rather than replacing, for a large number of occupations, meaning that new jobs will be created, whilst others jobs are likely to be altered, rather than simply displaced.
Big productivity gains are likely to be in sectors that provide benefits across the economy, such as manufacturing and agriculture.
Recent years have seen a revived interest in the importance of manufacturing across the democratic world. Part of this is the need for resilience highlighted by the pandemic. Part is caused by gradual decoupling. And part is based on the need for manufactured goods, notably semiconductors, to help power technological advance. I made the case for an increased focus on manufacturing in this BBC series called Rethink in 2000 and a number of governments around the globe have started to actively advocate and act for “reindustrialisation”.
What was lost with the decline of manufacturing in the West was nicely summed up by Paul Kennedy, born in Wallsend near Newcastle, and the author of the Rise and Fall of the Great Powers:
There was a deep satisfaction about making things… A deep satisfaction among all of those that had supplied the services, whether it was the local bankers with credit; whether it was the local design firms. When a ship was launched at Swan Hunter all the kids at the local school went to see the thing our fathers had put together… this notion of an integrated, productive economy was quite astonishing.
As well as sustained political momentum, there is strong public support for so-called “physical” industries like manufacturing and agriculture, with the public regarding them as producing important and meaningful economic and community contributions, as well as dignified, proud jobs. Importantly, the productivity and growth boost that AI will deliver could well bring particular boosts to sectors such as manufacturing.
AI can help manufacturers in were once tasks that constrained productivity growth, such as predicting when machines need maintenance; taking a more detailed approach to quality control; and identifying inventory blockages and supply-chain needs. It is also able to predict peaks and troughs in demand, meaning that there is no longer any need to make the false choice between efficiency and resilience. So-called “cobots” are also working alongside human workers in factories to help with more detailed or mundane tasks.
AI playing an important role in boosting manufacturing productivity is essential to the economy as a whole. Manufacturing tends to be much more R&D intensive and, historically, has a much bigger impact on productivity growth than services. An AI powered manufacturing revival is also likely to play an important part in creating secure jobs that don’t require a university degree and reviving areas that have been stagnant since deindustrialisation.
In other words, AI might play a part in not only creating jobs, but in creating just the kind of jobs that communities have lost in recent decades. The growth of AI, allied with other societal, economic and political pressures to reindustrialise and build resilience might herald a tech-driven industrial renaissance in large parts of the democratic world. And this means that AI will play a part in tackling the so-called “hourglass economy”.
AI might go some way towards tackling the hourglass economy, dominated by knowledge jobs at the top and insecure jobs at the bottom.
I’ve made the case in my two books, Little Platoons and The New Snobbery, that the UK and other Western states are characterised by what has become known as the “hourglass economy”. This is the division of the economy into well-paid, respected jobs at the top in the “knowledge economy” and poorly paid, insecure work at the bottom as part of the “precarious economy”. Michael Sandel argues that this has led to a division of esteem, based on “credentialism” and education levels.
The jobs that used to characterise the middle have largely disappeared, notably jobs in manufacturing, that were proud, dignified and secure and didn’t require university degrees. And this was defined by the so-called “left-behind” towns, with growing disengagement, alienation, growth in addiction and even “deaths of despair”. The economic divide is in many cases a regional one (coastal v heartland in the United States, North and South in the UK), and has often led to a broader political polarisation.
This divide isn’t an internet related change, but is connected to the decline of heavy industry in parts of the West. The relevance of tech is that many of the knowledge jobs at the top of the economy are, at least indirectly, tech related.
An important potential of AI is that it will not only boost productivity in the economy, but also recreate dignified work and rebuild the “hollowed out middle”. This idea was picked up recently by David Autor, an economist at MIT, who had previously been a fierce critic of tech. As the New York Times reports on Autor’s views:
A.I., if used well, can assist with restoring the middle-skill, middle-class heart of the U.S. labor market that has been hollowed out… It can… change the economics of high-stakes decision-making so more people can take on some of the work that is now the province of elite, and expensive, experts like doctors, lawyers, software engineers and college professors. And if more people, including those without college degrees, can do more valuable work, they should be paid more, lifting more workers into the middle class.
As Autor’s report puts it:
The unique opportunity that AI offers humanity is… to extend the relevance, reach and value of human expertise to a larger set of workers. Because artificial intelligence can weave information and rules with acquired experience to support decision-making, it can enable a larger set of workers equipped with necessary foundational training to perform higher-stakes decision-making tasks currently arrogated to elite experts... In essence, AI — used well — can assist with restoring the middle-skill, middle-class heart of the U.S. labor market… Artificial Intelligence is this inversion technology. By providing decision support in the form of real-time guidance and guardrails, AI could enable a larger set of workers possessing complementary knowledge to perform some of the higher-stakes decision-making tasks currently arrogated to elite experts... This would improve the quality of jobs for workers without college degrees, moderate earnings inequality, and — akin to what the Industrial Revolution did for consumer goods — lower the cost of key services such as healthcare, education and legal expertise.
For Autor, AI becomes a “worker-complementary technology”. This is borne out by research showing that AI can be a job complementing, rather than a job replacing technology. Research by economists at the MIT found that the biggest benefit from Ai came to lower-skilled workers, with AI allowed the skills of higher-skilled workers to be disseminated across the workforce more easily. This led them to argue that, “the lesson is that, more often than not, you’ll benefit by augmenting workers rather than trying to replace them.”
Unlike in previous waves of automation, governments and industry are serious about tackling skills needs.
Despite the reasons for optimism, it’s clear that AI will change the nature of work and the labour market, but all interested parties are taking the change much more seriously than at any other stage of automation. That, in itself, is a cause for optimism.
George Orwell once reportedly said that the problem for British coal mining was that there was no coal in Hyde Park. In other words, despite coal powering Britain’s industrial rise, the major coalfields were always both literally and mentally some distance from the corridors of power. This was something that was still reflected in the Brexit referendum when 41 of the 42 former coalfields voted to leave the EU.
The same could be said for previous waves of automation, which generally had a bigger impact on blue collar than white collar jobs. Previous waves were either regarded with a laissez-faire sense of inevitability or governments simply weren’t ready or prepared for the changes that automation brought. At no point during previous waves of automation or technological change did governments give sufficient consideration to the reskilling needs necessary to cushion local job displacement from automation.
The AI driven change is different. And that might partially because professional jobs are as likely to be impacted as manual jobs. This, obviously, concentrates the mind of the professional class who tend to dominate politics and (for obvious reasons) the upper echelons of business. Another important element is that the character, if not the precise form, of labour market change is clear. Most importantly, though, there is a clear determination on the part of governments, tech companies and civic society to do whatever possible to prepare societies and economies for the change and provide active support around reskilling, skills needs and sectoral consequences.
National governments are taking important steps to ensure that their populations have the necessary skills to succeed. SkillsFuture and the AI Skilling Fund in Singapore, the EU’s Skills Strategy for Europe and the UK’s AI Skills programme are all examples of the kind of public programmes with a scale and ambition that simply didn’t happen in previous waves of automation and tech-driven change. Similarly, the involvement of all of the major tech companies and AI labs in reskilling efforts is indicative that business also understands its role in helping workers to adjust. This can be seen with the recently launched AI-enabled Workforce Consortium and a number of measures, such as Google AI Career Certs and Microsoft’s Future of Work initiative.
What next? Four more steps to build on AI skills
There are early signs for optimism and early signs that governments, businesses and civic society are taking necessary steps to cushion the impact of a changing labour market.
Constantly monitoring how skills needs will change
One constant in the history of automation is that more jobs will be created, but also that skills needs will change. There’s no point in compiling one report about AI skills needs and then thinking that the job is complete. We know that AI models are developing at a remarkable speed. There’s little reason to think that AI related skills will somehow stand still in such a rapidly changing environment. Instead, it’s important to constantly focus on how skills needs might change. This might involve mapping emerging skills needs against an existing skills base; using tools such as scenarios to consider what this might mean in the longer-term; and building flexibility into the curriculum, so that it can pivot more quickly as needs change.
Ongoing skills development should become the norm
The advance of technology means that the so-called “half life” of a skill, meaning the amount of time it takes for a skill to be half as useful has dramatically reduced. Whilst it used to be a decade or more in some sectors, it is now as low as two years. Skills learned while at school or college are unlikely to be enough to keep workers’ skills relevant throughout their career. As such, both governments and businesses should work to ensure that ongoing upskilling is a possibility. This could involve using “skills wallets” on the SIngaporean model and governmental support to help ensure that smaller businesses don’t get left behind.
Developing AI Skills Institutes is at least as important as developing AI Safety Institutes
The Bletchley Park AI Safety Summit led to the creation of a number of national AI Safety Institutes, including in the UK, France, Japan and the United States. Britain and France have also announced an AI Safety Partnership. Building up an AI skills base is at least as important as countries look to make the most of the substantial opportunities offered by emerging technology. These institutions could be industry and government partnerships that provide good practice, emerging research and help ensure that countries are not eschewing the enormous opportunity of AI.
Partnerships between industry, governments, unions and business will be crucial
The age of AI must also be the age of partnerships between tech companies and all other sectors of the economy and society. Partnerships between tech and governments will be essential to ensure adequate knowledge sharing. Deeper relationships between tech and manufacturers will help physical industry effectively utilise AI to boost productivity and create jobs. Partnerships with labour unions and small business organisations will help ensure that workers and SMEs are able to adequately boost their skills.