The centuries-long shift to better jobs continues in the age of AI

Plus: Eurozone unemployment has never been lower, Cuba is running out of oil, and more

Feb 02, 2026

Welcome to The Update. I’m experimenting with the format to maximize the signal.

In today’s issue:

The centuries-long shift to better jobs continues in the age of AI

The period since the Industrial Revolution has seen a steady improvement in the quality of jobs. Not only have workers moved from agriculture to manufacturing to services – typically to better pay and less strenuous work – they have also upgraded within these sectors. For instance, secretary jobs have shrunk as a share of service employment in recent years, while consulting jobs have expanded.

Source: ‘Technological Disruption in the US Labor Market’

Recent predictions suggest that AI will put a stop to this trend. Many of them come with thinly veiled schadenfreude: finally, it’s the well-paid office workers who will suffer. But if so, we’re not seeing it yet. The Economist shows that the wage gap between white-collar and blue-collar workers continues to rise much as it has for many decades.

And within the white-collar category, the more advanced jobs fare better. Routine back office jobs are down, whereas jobs for managers, technical specialists, and other complex roles are up.

Claims that long-running automation trends were breaking were common even before the rise of LLMs. Keller Scholl and Robin Hanson tested these claims in 2019 and found them wanting: the types of US jobs most susceptible to automation didn’t change from 1999–2019. It’s true that AI may eventually change this, but near-term predictions of disruption often underweight historical trends. Patterns that have held up for hundreds of years typically reflect strong underlying forces.

While AI hasn’t replaced radiologists, it is improving cancer detection

Radiology is a good example of how we often overestimate how quickly advanced jobs will be automated. Geoffrey Hinton’s 2016 advice to stop training radiologists hasn’t aged well, as demand for them has continued to grow. But AI is, in fact, increasingly used in radiology – for instance, to read breast cancer scans. A Swedish randomized controlled trial found that replacing one of two radiologists with AI improved detection rates. Applying this across all screenings would reduce total radiologist workload at breast centers by 4.5 percent, according to a recent Norwegian study.

Mandatory retirement of professors should be about contribution

According to Harvard’s latest data, 37 percent of tenured professors at the Faculty of Arts and Sciences are over 65, and 44 percent of retirements occur at 75 or older. Many other universities without mandatory retirement also have high proportions of older faculty.

Should they be forced to retire to help younger researchers get a job? As Nature notes, many think so. Others find this ageist, arguing that older professors have a right to keep working.

I think both sides in this debate have the wrong perspective. Academia doesn’t exist to give jobs to academics, but to provide value for students and wider society. Consequently, older professors should only have to retire if they provide less value than their younger replacements.

Mark Rutte exaggerates Russian strength

A few months ago, NATO Secretary General Mark Rutte said that if Sweden doesn’t raise its defense budget to five percent of GDP, it should start teaching Russian in school because ‘you need to learn that language then’. Now he says that ‘if anyone thinks that the European Union or Europe as a whole can defend itself without the US, keep on dreaming’. I don’t think he believes these claims himself. Russia has struggled to defeat a country with a tenth of the EU’s population and a much smaller share of its GDP. You might argue that Rutte has good political reasons for what he says, but that doesn’t make it true.