There is much discussion at present about the benefits artificial intelligence (AI) might bring to society. We are told it will transform productivity, revolutionise employment, create new opportunities, and deliver economic growth on a scale we have supposedly never seen before.

None of those claims is proven. Many might be wrong. But much less discussed is the fact that AI is already imposing substantial costs on society, whether or not people use it. I suspect that most people are completely unaware of the extent to which they are already paying for the development of this technology, which might deliver them remarkably little in return.

AI and inflation

The most obvious cost is inflation. The extraordinary demand for computer chips, memory, and other electronic components created by the AI industry is already significantly affecting the prices of a wide range of products. Computers, mobile phones, gaming consoles, cameras and almost anything else that requires sophisticated electronic components are vulnerable to these pressures.

The AI industry is competing for manufacturing capacity and components that would otherwise be available to producers of consumer goods. When demand increases faster than supply, prices tend to rise. The consequence is that people who might never use an AI application are already paying more for products that they need.

This is, however, only one of the ways in which AI is creating inflationary pressure. Another is through its extraordinary demand for electricity. Data centres require large quantities of power, and the scale of investment now proposed means that this demand is likely to increase considerably. In an electricity market where prices respond to changes in supply and demand, the consequences are obvious.

If demand rises faster than generating capacity, prices can increase, and those increases are not confined to the companies operating data centres. They can affect every household and business connected to the electricity system. The result is that AI can increase the cost of living for people who have no interest whatsoever in using it.

There is another dimension to this problem. Some proposed data centres are considering building their own gas-fired electricity generation facilities because the existing electricity grid cannot provide the power that they require.

That might appear to solve the problem for the companies concerned, but it does not solve the problem for society. Instead, it increases demand for gas, which can put further upward pressure on energy prices. It also creates additional carbon emissions at precisely the time when we should be reducing our dependence on fossil fuels.

We are, therefore, facing the possibility that the AI industry will both increase the cost of energy and make the transition to a sustainable economy more difficult. That is a considerable price to pay for a technology whose wider benefits remain uncertain.

AI and distorted financial markets

The consequences of AI are not confined to the prices of goods and energy. The extraordinary amounts of money being invested in this industry are also having an impact on financial markets.

AI companies and those building the infrastructure that supports them are attracting enormous sums from investors. Some of that money comes from the sale of other financial assets, including government bonds.

When bonds are sold, their prices fall, and because bond prices and yields move in opposite directions, the interest rates implied by those bonds rise. There are, of course, other factors influencing bond markets, and the precise contribution made by AI investment is difficult to establish. Nonetheless, the possibility that AI is significantly contributing to upward pressure on bond yields deserves considerably more attention than it is receiving.

That's not least because the consequences of higher bond yields are significant. They can influence the cost of mortgages, rents, business borrowing, car finance and many other forms of credit. They can also increase the apparent cost of government borrowing, which politicians then use as an excuse to demand cuts in public spending.

We are constantly told that governments are responsible for higher interest rates because they have supposedly borrowed too much, spent irresponsibly - which is a claim that only appears to apply to social security spending - or failed to satisfy the demands of financial markets. What is much less often considered is whether the diversion of financial resources towards speculative investment in AI might itself be contributing to the pressures that governments are then blamed for creating.

If that is happening, the irony is considerable. Private financial speculation would be imposing costs on the whole economy, with government expected to accept responsibility for the consequences.

Other markets

There are also substantial physical costs arising from the AI investment boom. The construction of data centres requires skilled labour, electrical engineers, building materials, generating equipment and access to infrastructure. None of these resources is available in unlimited quantities.

If AI companies are willing to pay more to secure them, other projects will inevitably face increased costs or delays. That includes housing, where construction costs are already a serious problem, but it also includes essential infrastructure maintenance, transport projects and flood defences.

The consequence is that resources which could be used to improve the quality of life for millions of people might instead be diverted towards the construction of facilities designed to support AI. That is what economists call an opportunity cost, and it is a very real one.

Then there is the impact on employment. AI is already being used to replace some forms of work, and there is every reason to think that this process will continue.

What is particularly striking is that some of the jobs now threatened are precisely those that we were being told, until very recently, were essential to the future of the economy. We were encouraged to believe that increasing the number of people with IT and related technical skills was vital. Now some of those same people face the possibility that their work will be devalued or displaced by AI.

The consequences will not be confined to those who lose their jobs. There will also be downward pressure on wages, increased insecurity, reduced spending power and potentially greater inequality as the rewards from economic activity are redistributed away from those who work towards those who own the technology.

What are we getting in return?

All of this raises an obvious question, which is what society is actually gaining in exchange for the costs that AI is imposing upon it.

There are undoubtedly some useful applications of AI, including in scientific research, medicine and other specialist areas. I am not suggesting otherwise. But for the overwhelming majority of people, the benefits experienced so far appear to be remarkably modest.

Most people have marginally better web searches. We can produce pictures and images that nobody particularly needs. We can manipulate photographs more easily.

These are hardly insignificant developments in every case, but they do not obviously justify the enormous economic and environmental costs that are already being incurred.

What is particularly troubling is the way in which those costs are being distributed. The benefits of AI, if they eventually materialise on the scale promised, will largely accrue to the companies that own the technology and to their shareholders. The costs, however, are already being imposed on everyone else.

We pay through higher prices for electronic goods, potentially higher electricity bills, increased pressure on the cost of housing and infrastructure, greater employment insecurity and possibly higher interest rates.

Costs are being socialised: profits are being privatised

In other words, the costs are being socialised whilst the potential profits are being privatised. That is a familiar pattern in modern capitalism, but it is no more acceptable for being familiar.

The question that governments should be asking is not whether AI represents an exciting technological development. It undoubtedly does. The question is whether the resources being devoted to it are justified by the benefits that society can reasonably expect to receive.

That requires an assessment of the full economic, social and environmental costs of AI, including the costs that are currently being passed on to people who have no say in the development of this technology. It also requires governments to consider whether the AI industry should be required to bear those costs rather than being allowed to impose them on everyone else.

At present, there is remarkably little evidence that these questions are being seriously considered. Instead, we are being told that AI is inevitable, that it represents progress and that we must accommodate it whatever demands it makes upon our economy.

I do not accept that argument. Progress should be measured by the extent to which AI improves human well-being, not by the amount of money invested in technology or the market valuations of the companies promoting it.

If AI makes life more expensive, increases inequality, diverts resources from essential public investment and creates greater economic insecurity, then we have every right to ask whether it is delivering progress at all.

The costs of AI are not something that we might face at some point in the future. They are already being imposed upon us, and most people have no idea that they are paying them.

It is time we started asking whether AI is worth the price.

PDF of article