The Challenge Network

 go back 

An automated world.

An automated world.

There is much concern that automation, embodying some element of artificial intelligence, will displace the workforce. There is no evidence for this to be found in the productivity figures – indeed, zero correlation between IT intensity and productivity growth in an industry – but the meme is a strong one and needs discussion. This note considers the implications of these beliefs.

Background:

The emerging economies are set to supersede those of the old rich world; indeed, passed them in aggregate output in the early 2010s. They will field a young and educated workforce that represents three times the entire OECD population, and will have near-parity of access to technology. This points to formidable levels of competition in the 2030s.

One consequence of intense competition is cost convergence. Every firm studies its rivals segment by segment, in a process known as benchmarking. Best practice is copied and improved upon. Firms come to be more and more like each other, which is to say, to share a common cost and quality level. As the marginal costs of the least efficient producer define prices in a free market, this means that price falls universally to near cost. This is advantageous to consumers but a disaster for firms in the industry, from which profitability evaporates. This process is known as “commoditisation”, and is a universal. Consider the desktop computer, once built to a 60-70% gross margin, now supplied at a few percent.

Commoditisation is offset by innovation. New activities are innately inefficient and their cost curves are steep: profits are high. If they grow rapidly, they are able to invest strongly. Provided that growth in output exceeds the growth in productivity, they will take on labour. (Firms in the last phases of commoditisation suffer the opposite, and shed labour and capital as enforced productivity exceeds growth.) Economies proceed through this factor recycling, whereby resources used by industries are absorbed in young, inefficient sectors and shed in mature, highly efficient ones.

Enter automation

Automation has the potential to accelerate commoditisation, and to cut the factor recycling process. Firms will decide whether to automate by looking at the cost of the two options, weighted for risk. Labour-side risks involve industrial action and issues such as disease and labour-protecting regulation. Automation has the risk of dedication to systems that may go out of date, suffer from systematic weaknesses or lack trained staff to maintain them. Automated systems are inflexible and ill-adapted to the alternating demands from the marketplace to which human workers can adapt themselves. There is, therefore, an “eggs in one basket” risk that many firms fear; but automation offers quality control, access to finely tuned supply chains and above all, cost and functional predictability.

Automation is often seen as the direct replacement of workers with machinery, and certainly, some manufacturing – in vehicles, in electronics manufacture – where this is obviously so. In most activities, however, it is a matter of synergy, where access to the right equipment allows the worker to be much more productive, and enjoy the benefits of increased added value in terms of earnings. Consider a contract gardener, once arriving on a bicycle to use the house owner’s tools to fulfil tasks. This was slow and inefficient. Their modern equivalent arranges their day by cell phone, arrives in a pickup full of machinery and handles in an hour what his former colleague would have need half a day to handle. Farms that once needed twenty labourers are now handled by two or three people, and a host of machinery that those individuals spend most of their time maintaining. The modern economy is human and machine, working together, to generate more than either one can handle on their own. This has generated high levels of employment, with wages strongly skewed toward high skilled jobs that are particularly demanding of this synergy.

Enter AI

This synergy is challenged by hypothetical developments in computing, whereby the human element can be replaced by automation. Self-driving vehicles might displace twenty million jobs in the US alone, as everything from truckers to taxis drivers are replaced by software. To assess the broader reality of this, one needs to look closely at what is meant by artificial intelligence.

What people mean by “AI” today is one of two things: either dedicated hardware solutions, such as self-driving cars, or broader software systems rooted in an elderly technology called neural networks. More here. What the wild-eyed prophets of social doom mean is something different, so called general AI (gAI), whereby a self-aware machine (that is content to remain an effective slave) can perform any task better than a human. gAI is a long way off, as we have no idea how consciousness is generated. If it arises by accident (through so called emergence within sufficiently complex systems, perhaps how a mouse or a newly born human become aware) then its utility will be questionable. How can it be induced to take orders? What happens when it has mental problems, starts to save for its retirement, asserts its personhood and starts suing its former owners? The ethical and operational problems are extreme. Yet human level computation should be generally available in the 2030s for consumer-level prices, and experiments will be done. Some fear the self-propagating power of such outcomes – a grand gAI that will turn the world into Mordor with Terminators in place of Orcs – but this is fanciful, and better perspectives will need better understanding of what is involved in awareness. Perhaps extreme minds divide spontaneously into sub-minds that fracture from their parent, self-limiting the process. Without better technical insight, we have no idea about any of this.

A world with powerful AI in it

For the purposes of this exercise, let us suppose that human-engaged or supra-human AI can be deployed usefully in industry under the economic conditions that we have suggested will obtain in the 2030s. Anything which the contract gardener could do, it can do better, and Bob the Builder will indeed deploy thinking cement mixers and devices which 3D print entire houses, complete with plumbing and fixtures. What happens when there is no room for a human workforce?

The conditions of the 2030s will ensure that any form that does not immediately take up the best solutions available to it will be rapidly marginalized. At issue is what is meant by “best”, a contextual term set by state regulation, costs and other factors.

At best, human labour will be rapidly displaced, social disturbances will follow and states will take measures to maintain demand and quieten tensions. The differing responses will, undoubtedly, generate trade tensions, just as embodied pollution and other negatives within traded goods will do the same more promptly, probably in the next decade. Capital is then virtually worthless, and the financial system undergoes radical change. As inflation is tightly limited - no labour demand, endless supply - states can manage demand via the money supply. Positional goods - desirable things which cannot be replicated, such as properties in good locations, antiques and so on - undergo major asset price inflation.. Sources of defensible competence - intellectual property, anything which conveys excess profitability, a so-called 'rent' - are particularly bid up, and the owners of such assets become hugely wealthy, and accordingly taxed. Pension assets, however, become close to worthless and the state will have to pick this up. The value of a state's currency is defined by its balance of payments - demand for the currency - and its long term prospects, itself predicated on social peace and innovation. City governance - the key element of governance in this world of cities - sets regulatory criteria for innovation amongst its participating industries.

It is not at all predetermined that such sweeping impacts will be generated, however. It is far more likely that the synergy aspect of human and technology will lead to a considerable economic boom. Consider a company. It is an abstract legal entity that transduced a wide range of signals and flows into similar outputs, ad harvests added value from the process. It is, if not an organism, then an AI, running on a machine made of people. Since its inception prior to the Industrial Revolution, the joint stock company has been an extraordinary engine of growth, displacing the informal craft based system so the previous centuries. Can we expect less of information augmentation of this familiar entity?

In order to explore the outer edge of this future, however, let’s stick with the notion that independent machinery can do anything that humans can, only better and cheaper than they.

Economics

Utopians see this as a world of free goods, universal income from the state and general bliss. However, let’s revisit the notion of commoditisation.

As noted, this is anyway a world of extreme competition. Firms are forced to converge on best practice, which trims out labour and so reduces costs by 30-40%. Prices fall to the marginal cost of the least efficient producer, which as we have seen, crashes industrial profitability. Returns to capital thus approach zero, despite massive cost reduction. The incentive to invest, save to avoid being destroyed by the competition, evaporates.

So, too do both dividends and wages. National income is confronted by a paradox: there is a cornucopia of high quality, low cost goods on offer but there is no money in circulation with which to buy them. However, the inflationary impulse is also gone, as there are essentially no supply limitations, and there is certainly no labour drive to raise wages. Market forces are unable to demand high dividends. States can, therefore, print money with impunity, save through the foreign exchange mechanism. Currency valuation becomes all –important as a determinant of national well being, set by the long run prospects of the economy – investment, innovation and political tranquillity – and by the acute demand for the currency, set by public national borrowing, debt repayment and the general the balance of payments.

Social and political pressures are high to generate demand. The social contract must be maintained. Nation states either allocate production for consumption - major infrastructure projects, welfare make-work and so on, or more generally print money to support a market mechanism. The crucial concern is to keep investment flowing and high end technology used for innovation, so as to secure long run competitive position. This exacerbates the problems of commoditisation, and the self-reinforcing cycle becomes embedded.

There is no reason for anyone to invest in poor countries, as their labour is not needed and their demand feeble, so development comes to a halt. Charitable activities by rich countries would seek to redress this balance, with the most modest of success.

The “feel” of the 2040s.

The 2040s is a time of stark environmental limitations, with 80% of the world’s population stacked up in over a thousand mega cities. Managing these brittle social, economic and environmental systems is a central priority, and AI-augmented government would focus unwavering attention on this. (In the absence of AI, human systems perfected in the C20th would keep a similar watch.) Life is lived in total transparency, with your life managed “for your own good”, and more or less gently kept on officially-sanctioned tracks. A central question of the time is what to do with the surplus population, defined as the portion which costs more to keep than it adds in value. In an extreme AI world, that is most of the population, in a less automated one it is a considerable fraction of it. In the Western economies as of today, roughly half of the population are net beneficiaries of state welfare, in the sense of paying less tax than they receive in cash and kind. admittedly, many of them work, which they will not do in the extremes of the 2040s. The extent to which future states will mandate workfare – tasks that require a human face, such as visiting the many elderly of the time – and the extent to which hobbies and pastimes such as gaming and life in virtuality will be encouraged is anyone’s guess, but will vary amongst the polities of the world.

So the world of 2040 will be very different from today’s societies. The extent to which automation will modulate that difference is clearly significant, but prescribes a way in which a greater or lesser fraction of the world’s population will be managed. It seems a pretty hopeless end game, unless gAI provides a scape into which individual consciousness can escape, to growth and exploration. This is what followers of the Singularity anticipate, but it might be advisable to take that with a whole table spoon and not a pinch of salt.