And yet I have reassessed my own outlook. As an economist, my view in recent years was that groundbreaking innovations in AI should be a net positive, because that has been the case with the arrival of new technologies throughout my career. Reality never vindicated arguments by techno-skeptics and pessimists that some new technology would create a crisis of mass unemployment, so my default view was that this time should be no different. Moreover, AI could be a godsend for Western societies confronting problems such as a rapidly aging workforce, political backlashes against immigration, and rising government debt burdens.
But the response I often hear from those most deeply immersed in the field is that I don’t “get” how frighteningly fast AI development is proceeding. I have failed to appreciate just how dangerous the technology could be, especially if it reaches the point where it can adapt its own thinking without any input from its controller. By the same token, if AI ends up in the wrong hands – be it criminal organizations or rogue states – it could cause political and social havoc.
And then there is the extraordinary cost of producing and operating this technology. We just watched US President Donald Trump visit the United Kingdom, and alongside all the pomp and ceremony was an intense focus on mobilizing even more investment in data centers and the energy infrastructure to power them.
Given the unprecedented market valuations and the sheer scale of the sums involved, I must ask: What, exactly, is the purpose? If AI, with all its obvious power, cannot be marshaled in ways that materially solve our biggest problems, what will all this have been for? I can see why the owners of the leading companies want all other businesses to adopt AI. But that does not mean the technology will help us address the biggest, most persistent challenges that our societies face.
For private-sector employers, AI does seem like an obvious way to keep labor costs in check. Most economists would welcome anything that can shock corporate executives out of focusing so much on the need for cheap, flexible labor. But this issue is not especially important in the big picture.
In the UK, for example, productivity growth has been weak ever since the 2008 financial crisis, and if you dig into the data, you find that key public-sector institutions are a major factor. Yet while everyone talks about boosting productivity at the National Health Service, no one has really made it a high priority.
When I served on the independent Times Health Commission in 2023, I often had to push back on the widely shared view that, regardless of what we might recommend, health spending as a share of GDP would inevitably continue to rise. But why should this be the case? In the late 1970s, health spending as a share of GDP was about the same as education – around 5%. But today it is closer to 12% of GDP, more than double that of education.
To my economist’s eye, these trend lines make no sense. If anything, education spending ought to be higher than health-care spending, because better educated young people can both boost the productivity of the future workforce and make better health choices. It may also be easier to achieve higher productivity in health care than in teaching. Making matters worse, the more that spending on health care rises, the more difficult it is to avoid spending cuts in other areas, barring more large tax hikes. The UK Office for Budget Responsibility continues to project horrendous levels of potential government debt by 2050 unless we do something about the NHS, welfare payments, public pensions, and other social services.
So, my recommendation to public policymakers, industry leaders, and others is to develop a serious, clear, evidence-based plan to ensure that AI contributes to productivity growth where it is needed most. It won’t necessarily be easy, but with all the brain power and talent being deployed to develop the technology, surely someone can figure out how to put it to good use.
