Artificial intelligence, work and doomsday scenarios. The apocalypse postponed

Whilst we await the next apocalypses brought about by artificial intelligence, we can note with a certain degree of calm that at least one has been postponed: that of employment. We were expecting armies of the unemployed, young graduates rendered redundant by ChatGPT, office workers replaced by bots, and programmers swept aside by artificial programmers. It may still happen. But for now, it really isn’t happening.

The US data compiled by The Economist is almost infuriating for those who had already begun to prepare for the economy’s demise. In August, the US economy created 162,000 jobs; the unemployment rate stands at 4.1 per cent and is lower than that recorded in 90 per cent of the months over the last fifty years. Even young people aged between 20 and 24 – cited as the first victims of generative automation – show an unemployment gap relative to the average that is close to the lows of recent decades.

The Economist’s estimate is striking: around one million jobs created so far by artificial intelligence, compared with around 200,000 redundancies attributed to AI since mid-2023. Roughly five to one. This does not mean that every person made redundant has found a better job just round the corner. It does mean, however, that the first overall picture of AI does not resemble a machine that takes away jobs: it resembles a technology that shifts them, transforms them and, at least so far, generates new ones.

The most interesting part of the story is understanding where these jobs are being created. AI lives in the cloud, but the cloud requires concrete, copper, turbines, transformers, air conditioning units, electricity grids and technicians. Compared with 2022, annual US expenditure on chips, servers, data centres, cooling and energy is expected to have risen by around $500 billion. In the five sectors most closely linked to this infrastructure, The Economist estimates around 320,000 additional jobs compared with the trend. Electricians, fitters, HVAC technicians and network engineers are needed. When a technology boosts the demand for electricians, perhaps it is worth revising the notion that AI is only about nerds and algorithms.

Then there is the paradox. Some jobs that were supposed to be wiped out by AI are actually growing. Between 2023 and 2025, employment of paralegals rose by around 11 per cent, whilst that of market analysts rose by 6 per cent. If a technology enables people to work faster, costs fall and demand may rise. Productivity creates a market.

But this is precisely where the way we discuss technology needs to change. The relevant question is no longer: ‘Will AI destroy jobs?’. Of course it will. Every major innovation does. The question is: how many will it destroy, how many will it create, which wages will it boost, which skills will it make more valuable, and how quickly will we be able to move workers from declining jobs to those that are emerging?

For two years we’ve had prophecies and very little data. Now the data is starting to come in. It doesn’t prove that the apocalypse will never come. It proves something simpler: it hasn’t arrived yet. And whilst we were waiting for it, someone started hiring.