Why AI Has Not Resulted in Widespread Unemployment

People have become used to living with AI fairly quickly. ChatGPT is barely three years old, but has changed the way many of us communicate or deal with large amounts of information.

It has also led to serious concerns about jobs. For if machines become better than people at reading complex legal texts, or translating languages, or presenting arguments, won’t those old fashioned human employees become irrelevant? Surely mass unemployment is on the horizon?

Yet, when we look at the big numbers of the economy, this is not what’s happening.

Unemployment in the EU is at a historical low of around 6%, half the level of ten years ago. In the UK, it is even lower, at 5.1%, roughly the level of the booming early 2000s, and it is even lower again (4.4%) in the US.

The reason why there are still so many jobs is that while technology does make some human enterprise obsolete, it also creates new kinds of work to be done.

It’s happened before. In 1800 for example, around a third of British workers were farmers. Now the proportion working in agriculture is around 1%.

The automation of agriculture allowed the country to be a leader in the industrial revolution.

Or more recently, after the first ATM in the world was unveiled by Barclays in London in 1967, there were fears that staff at high street bank branches would disappear.

The opposite turned out to be the case. In the US, over the 30-year period of ATM growth, the number of bank tellers actually increased by 10%. ATMs made it cheaper to open bank branches (because they needed fewer tellers) and more communities gained access to financial services.

Only now, with a bank on every phone, is the number of high street bank staff in steep decline.

An imposition?

But yes, AI will take away some jobs. A third of Americans worry they will lose theirs to AI, and many of them will be right.

But since the industrial revolution, the world has seen a flow of innovations, sustaining an unprecedented exponential economic growth.

AI, like the computer, the internet, the railways, or electric appliances, is a slow revolution. It will gradually change habits, but in doing so, provide opportunities for new businesses to emerge.

And just as there has been no immediate AI boom when it comes to economic growth, there is no immediate shift in employment. What we see instead are largely firms using AI as an excuse for standard job cutting exercises. This then leads to a different question about how AI will change how meaningful our jobs are and how much money we earn.

With technology, it can go either way.

Bank tellers became more valuable with the arrival of ATMs because instead of just counting money, they could offer advice. And in 2016, Geoff Hinton, a major figure in the development of of AI, recommended that the world “should stop training radiologists” because robots were getting better than humans at analysing images.

Ten years later, demand for radiologists in the US is at a record high. Using AI to analyse images has made the job more valuable, not less, because radiologists can treat more patients (most of whom probably want to deal with a human)

So as a worker, what you want to find is a job where the machines make you more productive – not one where you become a servant to the machines.

Any inequality?

Another question raised by AI is whether it will reduce or increase the inequality between workers.

At first, many thought that allowing everyone to access an AI assistant with skills in processing information or clear communication would decrease earning inequality. But other recent research found the opposite, with highly skilled entrepreneurs gaining the most from having access to AI support.

One reason for this is that taking advice is itself a skill. In my own research with colleagues, we found that giving chess players top-quality advice does little to close the gap between the best and the worst – because lower-ability players were less likely to follow high-quality advice.

And perhaps that’s the biggest risk AI brings. That some people benefit from it much more than others.

In that situation, there might be one group which uses AI to manage their everyday lives, but find themselves stuck in low-productivity jobs with no prospect of a decent salary. And another smaller group of privileged, well-educated workers who thrive by controlling the machines and the wealth they create.

Every technological revolution in history has made the world richer, healthier and more comfortable. But transitions are always hard. What matters next is how societies can help everyone to be the boss of the machines – not their servants.