Whenever discussing my book with a wider audience I often start with an attempt to find common ground. Almost always that is the admission that we are living through an era of heightened crisis, something which unsurprisingly meets with little disagreement. In FALC I refer to this as ‘The Great Disorder’, as climate systems breakdown, an eroding neoliberal order, demographic ageing and various other challenges converge. My own diagnosis is that, taken together, these present an existential threat to capitalism. The critical question being whether it is replaced by something better or worse.

Alongside these more frequently-cited crises are the economic implications of robotics, automation and AI. While technological change is often thought of as a physical, materially-embodied process, over the coming years it won’t be underpinned by advances immediately obvious as such, and as deep and machine learning diffuse across the economy the outward appearance of daily life will remain much the same. While films like Terminator, Blade Runner and Ex Machina haveshaped the popular imagination as to what the future might bring, for now it is far easier (and more profitable) to build deep learning solutions than advanced robotics. Although that is a long way short of human-like machines, or human-like machine intelligence, that doesn’t mean its effects won’t be transformational.

Machine learning is a sub-branch of AI concerned with the creation of computer systems able to perform tasks based on patterns and inference rather than preset instructions. Over the next two decades its application in an increasing number of fields will particularly impact middle class professions like law, accountancy and aspects of healthcare and education. Machine learning will do to these industries what the tractor and combine harvester did to agriculture, leading to a historic increase in productivity and simultaneous reduction in the need for human labour. In FALC I refer to this prior transition in agriculture as ‘peak horse’, with the Industrial Revolution heralding a shift from a world where 50% of the labour force was agrarian to one where it was little as 1% (like the United States today). Similar changes with machine learning, although a long way from the sci-fi predictions of ‘the singularity’, will lead to peak human. While this will not eliminate human labour from most industries, it will reduce it in a manner analogous to agriculture over the last two centuries. Importantly, machine learning solutions – unlike expensive agricultural equipment – takes little time to diffuse.

But while it is the middle class professions of the Global North which are set to confront the greatest problems, the biggest challenge of all may be the implications AI has for inequality not within, but between, nations. For more than half a century the global economic model has offered poorer countries one advantage: their young, cheap workforces could be leveraged for export-led growth. That powered the rise of China, South Korea and Malaysia as they first manufactured goods, grew industrial expertise and gradually climbed up value chains. When I was a child ‘Made in China’ was associated with shoddy plastic toys. Today it is the world leader in fields as diverse as synthetic biology, mobile payment systems and high speed rail. More importantly, it is one of only two ‘AI superpowers’. It’s been the most rapid and remarkable national transformation in history.

Until recently the presumption was that the same dynamic, albeit on a smaller scale, would apply to countries like Pakistan, Indonesia and Nigeria. These nations would benefit from Chinese wages rising just as East Asia did after the 1970s at the cost of Western European and North American workers.

Except we now know that won’t happen with higher automation altering the trajectory of development. Rather than capital pursuing the ‘spatial fix’, seeking cheaper labour elsewhere to ensure a higher rate of return, it is turning to technology instead. Goods produced in China today will either be made there tomorrow, only with lower levels of human labour, or ‘re-shored’ closer to their home markets. Some countries will continue to benefit from the old trajectory, like Bangladesh with labour-intensive textiles, but it’s no longer the default and will be significantly limited.

For poorer countries with growing populations this presents a problem. Take Nigeria: with a population of 201 million, expected to double by 2050, such changes – allied to issues of water scarcity and declining crop yields as a result of climate change – could herald economic and social breakdown. Contrary to the claims of today’s techno-optimists I doubt several hundred million people will see rising living standards as they join the ranks of service workers in personal training and interior design.

Attendant with the end of development based on cheap manufactured exports, there is also the problem of just how limited the rewards of artificial intelligence will be shared. PriceWaterhouseCoopers predicts AI will add $15.7 trillion to the global economy by 2030, 70% of which will go to China and the United States. While the extent to which such technologies could eliminate jobs is contested (FALC cites estimates ranging from the extreme to the conservative) even if there is little net disturbance regarding jobs it will exercise profound implications for uneven development. In this world the United States, China, and to a lesser extent the EU and Japan, would leave the rest of the world behind in a manner similar to that of the industrialising countries after the early 19th century. If anything it would likely be worse.

With manufacturing and services increasingly performed by businesses whose intelligent machines are based in the AI superpowers of the United States and China, it will be the case that Silicon Valley and Zhongguancun aren’t just be responsible for building social media platforms, but payment systems, accountancy and legal services, health diagnostics and autonomous vehicles. While it is plausible that Europe could at least attempt to catch up, while protecting its domestic market for foreign competition, for the world’s poorest poor it will be a very different story. Unlike with the manufactured exports of yesteryear, much of sub-saharan Africa, Central and South Asia, the Middle East and Latin America will be effectively closed off – the dependency and under-development of the Global South not only extended but exacerbated. 

This is already in evidence with what Kai-Fu Lee describes as the ‘7 Giants’ of AI research: Google, Amazon, Facebook, Microsoft, Baidu, AliBaba and Tencent. The four US companies are worth north of $3 trillion – far larger than the UK economy, while their Chinese counter-parts have a cumulative value of more than a £1 trillion. While these companies are already among the most valuable in the world, their respective advantages means they stand to become far larger still, making life difficult for legacy businesses in industries they wish to enter: Amazon with offline shopping; Alphabet with autonomous vehicles and smart home products; Tencent, Facebook and AliBaba with payment systems and banking (1.7 billion people worldwide don’t have a bank account). The march of machine learning likely means that these already huge corporations will play an ever expanding role in our lives, with US firms dominating markets in Europe and India, and China in East Asia and the rest of the Global South. Any business or public institution wishing to benefit from machine learning over the next two decades may come to depend on technology from one of these countries with the cost of entry, in terms of data and resources, simply too high for anyone else. 

Alongside a new divide between wealthy and poorer countries, or more specifically the AI superpowers versus the rest, there is also the issue of inequality within nations. While it is true that new jobs will be created, these will be in areas which retain ‘uniquely human skills’ – think anything which requires motor-sensory coordination like cleaning, care work or repairs and maintenance. Because of demographic ageing the care sector is already creating millions of new jobs, with 8 of the 10 fastest growing positions in America to be found in the care industry.

But given the average worker in the sector already earns as little as $22,000, the possibility of large numbers entering it will only serve to suppress wages further and increase in-work poverty. It is particularly difficult to see how such a dynamic might be managed within the AI superpowers, as an ever larger mass of working poor see the dividend of AI redounding almost exclusively to a tiny elite, making the America of The Great Gatsby an exercise in egalitarian restraint. The idea that progressives can offer incremental solutions in the face of this is absurd. As with so much else it necessitates a systemic response. If the left doesn’t offer it, the nativist right will.

Indeed this process is already underway, despite machine learning having barely arrived. The last two decades, and with it the emergence of a true digital economy, has seen a massive concentration of wealth and power, with the platform model inevitably tending to monopoly. In the United States 75% of venture capital funding goes to just three states: California, New York and Massachusetts, with 50% alone going to the Sunshine State. Four of the world’s ten richest people are involved in the 7 AI Giants, with Jeff Bezos personally worth $159 billion. By the same token regional inequality is intensifying in places as different as Britain, China, Germany and Mexico. That is not to say the digital economy is to blame, local factors are to be considered as well as the asset values of large cities and younger workers preferring the metropolis, but the diffusion of machine learning applications will only exacerbate this. The fact that one of the world’s most innovative car manufacturers, Tesla, is based in Silicon Valley – as is much of the new wave of space transportation companies, is an omen of the future. As information becomes an increasingly central factor of production, areas with AI as their comparative advantage will pull away from the rest. It will be the equivalent of offering electricity with your products when nobody else can.

The diffusion of AI, and thereafter robotics, will intensify the status quo of ‘winner-takes-all’ . It will create unprecedented global inequalities and lead to a progressively larger ‘unnecessariat’, primarily located in the Global South. And that’s all before touching upon the existential risks more widely discussed in the media. AI, even unfolding in a ‘business as usual’ manner, could be on a par with demographic ageing over the next 25 years as a social and economic challenge. Rather than creating solutions policy-makers, especially those beyond China and the United States, are unlikely to even be aware of the problem.