Elon Musk is not most people’s idea of a classic technophobe, so when the owner of Twitter warns of the dangers of artificial intelligence, it is worth sitting up and taking notice. Fearful that a new generation of ever-smarter machines threatens life on Earth as we know it, Musk was one of many at the cutting edge of technological change calling for a six-month timeout in the training of new AI systems.
There is nothing new in the idea that the machines are coming, and they are out to get us. Techno-optimists are right to say that the same arguments were aired by Luddites in the early 19th century. By this token, the chatbot ChatGPT is to the fourth industrial revolution what the spinning jenny was to the first – a product that symbolises the dawning of a new era.
In the past, there has been a pattern to events. New technology has arrived on the scene and has offered the prospect of doing things quicker and better. Fears have been raised of mass unemployment as machines take the jobs previously done by humans. Eventually, the pessimists have been proved wrong, and the new technology has led to higher levels of employment.
There is little doubt that AI will be a gamechanger that can bring to an end a long period of weak productivity and low growth common to western economies since the global financial crisis erupted 15 years ago. As was the case when tractors took the place of farm hands, a single machine will be able to do what countless workers formerly did. That’s really not in question.
What is in doubt is who will benefit from the boost to productivity. What if all the gains are seized by a handful of tech giants? What if history fails to repeat itself, and AI destroys more jobs than it creates? What if AI does lead to a net increase in employment, but the new jobs are less well-paid than the old ones? Put simply, what if it is different this time? That may well be the case.
Much of the debate around the impact of AI is based on conjecture. There have been studies galore that have sought to estimate the number of jobs that will be affected – potentially running into the hundreds of millions globally – but nobody knows for sure. That said, certain conclusions can be drawn with a reasonable degree of confidence.
One is that the pace of technological advance will not slow down, and will probably continue to accelerate. ChatGPT was launched last November, and by March, a new version was available. The call by Musk et al for a six-month moratorium has to be seen in the context of the geopolitical struggle between the US and China. Neither superpower wants to give the other the opportunity to forge ahead. The chances of Washington and Beijing getting together and agreeing to a joint pause seem remote.
Despite the speed at which the technology is advancing, a second conclusion is that there will be no immediate root-and-branch transformation of economies. Machines are expensive and workers cheap. Moreover, companies have invested heavily in their existing systems, and these sunk costs mean it will take time for the impact of AI to show up in investment, jobs and productivity figures.
However, once the change does occur it is likely to be highly disruptive, because whole swaths of middle-class, white-collar jobs are at risk. This will be a break with the past, when earlier waves of technological advances made it possible for workers displaced from low-paid jobs to find better-paid employment in the new jobs created. People who were no longer needed as farm hands found work in factories.
AI poses a challenge to this model because of Moravec’s paradox – the notion that for robots, the hard problems are easy and the easy problems hard. Machines can wipe the floor with chess grandmasters, but have more trouble removing and cleaning the pieces at the end of the game: tasks that involve mobility and perception skills that have evolved in humans over millions of years.
But jobs that involve empathy and basic motor skills – social care work, for instance – tend to be poorly remunerated. That suggests the jobs most at risk from AI are likely to be higher paid than the ones created. There will be a boost to productivity and growth from the increased use of AI, but as things stand, the gains will be highly concentrated.
The final conclusion is that policymakers need to use the limited time available to them to respond to the obvious challenges. AI has the potential to bring great benefits, but also has risks that go beyond economics into the realms of privacy and ethics.
Launching the government’s white paper last week, the science and technology secretary Michelle Donelan said she wanted AI to be used to make the UK the “smarter, healthier and happier place to live and work”. All of which sounds marvellous, with echoes of an essay John Maynard Keynes wrote in 1930, predicting that within 100 years, increased prosperity would allow people to work 15-hour weeks.
Keynes’s vision has yet to materialise, and nor will Donelan’s unless urgent attention is paid to the 3Rs of AI: a system of global regulation that sets common standards for the use and development of AI; retraining to prepare the workforce for the inevitable change; and redistribution to ensure that the economic benefits are spread around. As with the climate crisis, the other existential threat of our time, the clock is ticking.
Larry Elliott is the Guardian’s economics editor
