I started reading this book just as the UK Government announced the setting up of an AI Economics Research centre chaired by Nobel Prize laureate Simon Johnson. So what is so special that merits a whole new field of study, and specifically for economists?

We have of course whole areas of economics that require a good understanding of specialist subjects such as health economics, behavioural economics, monetary and development economics, industrial economics and more.  And economists often specialise, as Andy Ross, Ian Harwood, Alvin Birdi and I explained in our book “How to be a Successful Economist” by becoming experts in one or more of those areas. But AI is all pervasive as it straddles both micro and macro and, as such, has something to say on all areas of economic activity. But is it good or bad, particularly for productivity? The truth is, for the moment, that we just don’t know. 

Why is that? Many reasons, as Ashot Davoyan tells us in this no-nonsense book. The difficulty arises because we can’t actually see and therefore measure AI’s direct impact, except possibly in SMEs anxious to cut costs, thus replacing the odd worker with greater use of technology. But when it comes to the wider impact he quotes Acemoglu who argued that it is innovations that are complementary to automation which in fact contribute to the wider economic gain. And it is diffusion that counts. 

How fast diffusion happens therefore matters. But how do these gains in fact register? One can see that the improvement is visible in some sectors already, particularly from US evidence where the adoption of technology in general and AI in particular has been faster than in most other countries and regions. But for the moment the overall economy-wide productivity improvement is eluding us in GDP statistics. It is partly because it is difficult to measure in the standard ways used by economists, as much of the AI activity, as Diane Coyle and others have argued, is ’invisible’. And for the economy as a whole the problem is one of aggregation, as the overall rise in total factor productivity is the result of not easily identifiable incremental productivity improvements by the public sector and by private sector companies, big and small. 

Instead, we focus upon potential  job losses across the spectrum, worry about the surge of  energy demand from data centres, fear an AI bubble and hear warnings by the Bank of England about a possible financial crisis through increased chances of cyberattacks on the financial system that the new technology may be creating.

Moreover, the huge capital expenditure required to keep up with demand is distorting debt and equity markets which are witnessing the emergence of a small band of mega, now trillion dollar firms.

But that as it may, the AI ”revolution’ seems unstoppable. We have already seen it affecting demand for labour, altering earlier work patterns both in industry and services, impacting on competitiveness and inequality and causing a rethink of what are viable economic structures by firms and governments. 

It could be said of course that we have been here before. Yes, there have been many technological revolutions that have left a profound imprint on the economy with certain jobs disappearing for ever. But as this book describes, improvements in accuracy with ” intelligent routing” and a reduction of what was perceived for a while as AI’s “hallucinations and harmful outputs” as well as ease of access, have seen the use of artificial intelligence spreading across the economy and to individuals a lot more widely than other innovations, impacting almost every area of society. 

The challenges for policymakers are enormous - and not only in terms of limiting harmful societal effects - see attempts at banning social media for schoolkids to save them from the addictive algorithms used by various platforms. How do you tax, control, limit access and, as capital spending by AI firms rises exponentially, stay on top of what comes next? 

And the question remains: will the AI revolution end up going dangerously out of control as many now fear? One can’t help but worry that our guardrails in this area, both economic and regulatory are rather weak. For these to be effective requires a level of international cross- border cooperation that is currently lacking, and sadly not just in the AI sphere.