AI should help not hinder employees – giving people more time to do the things they are best at

Is artificial intelligence a job creator or a job destroyer? Next year will, I suggest, give us a real feeling as to how AI will change the workplace – and the answer will be mostly good news.

Why next year? Well, start with three stories of recent days. One was that the head of AI at Google, Jerome Pesenti, predicted that the technology will start to hit a wall. His core point is that while AI is very good at doing specific tasks, it cannot replicate human intelligence: “We are very, very far from human intelligence, and there are some criticisms that are valid: it can propagate human biases, it’s not easy to explain, it doesn’t have common sense, it’s more on the level of pattern matching than robust semantic understanding.”

As for hitting a wall, he noted that the amount of power needed to solve broader tasks was making AI uneconomic. He did not quite put it in these terms, but I think his message is that human beings are better and cheaper.

The second story came from two longevity summits held at King’s College in London last month. Tina Woods, who is the chief executive of Longevity International, put the challenge of ageing populations this way: “Unless drastic action is taken,” she said, “people will suffer lower-quality wellbeing during a longer lifespan. This could become the next global crisis after climate change.”

Actually the crisis is more likely to run parallel to climate change, because it too is happening now. AI helps in a huge number of areas because what it is really good at is finding patterns and connections in data. So we are able to learn much more about what keeps people healthy and active as they get older. The more we know about the social determinants of health and the relationship between health and inequality, the better we can frame policies to help improve both.

The third story was a report by Jeffrey Funk, a former professor at the National University of Singapore, suggesting that AI was likely to be much less disruptive than some of its enthusiasts have predicted. He argues that there are indeed huge advances in AI software that will lift the quality of output of many white-collar workers, but this will take time.

“Viewed over the span of decades,” he writes, “the value of such software is impressive, bringing huge gains in productivity for engineers, accountants, lawyers, architects, journalists and others – gains that enabled some of these professionals (particularly engineers) to enrich the global economy in countless ways. Such advances will no doubt continue with the aid of machine learning and other forms of AI. But they are unlikely to be nearly as disruptive – for companies, for workers or for the economy as a whole – as many observers have been arguing.”

Those are just three snapshots. I think the message that is behind these and other stories comes in two parts.

One is that the world is getting much better at figuring out what AI is good at what it is not. It is wonderful at recognising patterns in a huge mass of data. So in medical diagnosis it is a dream technology, doing what humans do in a fraction of the time and cost, and doing it better.

For example, it is getting better at facial recognition, and the limits there will be social and political rather than technical. The Chinese are leading here, and many people in the west have reservations about a surveillance state. But most of us are happy enough to look at the camera at passport gates at airports, which of course use this technology. And if better use of cameras can make our streets safer, then most of us would say that would be OK too.

The other point is that getting AI to do the drudge work frees up human beings to be what they are better at, especially in that vast range of human interactions that fill our daily lives. If you had to pick out one single economic challenge, and I know there are lots of those, I would say is it increasing productivity in service industries. We know a lot about increasing productivity in manufacturing, but we struggle to replicate the 3 per cent a year that we get in manufacturing in service industries. If, however, you can get AI to do the more routine elements of a service industry – for example, by cutting the admin workload of teachers – then that frees up time to do the rest of the job better. You don’t have to look hard to see service industries, especially health and education, where resources are stretched because of the burden of routine (but necessary) record-keeping.

The moral of all this seems to me to be that service industry jobs should be made better by AI, rather than workers being replaced by a machine. Of course, most tasks will evolve, and some will disappear. But others will emerge. After all, there weren’t many AI engineers a decade ago, were there?