In an essay published on 26 August, Bill Gates argues that the current transition differs from previous technological revolutions. Earlier upheavals, such as the shift from agricultural to office work, unfolded over generations and created new categories of jobs requiring human judgement. In his view, AI is directly replacing cognition, in law, medicine, customer service and software, all at once.
This argument deserves to be taken seriously, not least because it comes from someone who has spent his career defending the opposite view.
Why this position matters
The reassuring argument everyone has repeated for three years is this: every technological revolution has destroyed jobs and created more, so this one will do the same.
That argument is historically accurate, and that is what makes it powerful. Today's jobs were unimaginable a century ago.
What Gates is challenging is not the history, but the transposition. Previous transitions shifted physical effort towards tasks requiring judgement. There was always a floor above to retreat to. If technology goes directly after that judgement, the floor above becomes less obvious.
What can be said against it
Three serious objections, and they need to be raised.
The data does not show this. Yesterday we reported on a survey in which more than 90% of executives see no effect on employment. A thesis of mass substitution should already be leaving measurable traces.
Full automation remains rare. The study of fifteen million interactions showed that less than 10% of uses fully replace a task. The dominant model is assistance, not substitution.
Job creation is invisible in advance. This is the strongest objection. By definition, you cannot name jobs that do not yet exist. Every generation has believed this time would be different, precisely because it could not see what came next.
Even if new jobs do appear, there remains an argument the objections above do not address: speed. The shift from agriculture to office work spanned three generations, allowing each one to adapt within its own career. A transition that plays out in ten years demands that mid-career adults retrain, which is a very different thing. The debate over the net number of jobs often masks the more important one about pace and who absorbs the shock.
What the debate is missing
There is one aspect these discussions barely touch on, which we addressed in our article on jobs.
The question is not just how many jobs, but which ones and for whom. A technology that automates the entry-level tasks of a profession does not eliminate that profession: it eliminates the way in. A young lawyer, a young developer, a young journalist used to learn their trade by doing precisely what AI now does.
This is what we noted about learning to code: when the tool produces the result, the path that trained people disappears. A profession can remain large while becoming inaccessible to beginners.
What to take away
We are not taking sides on a question no one knows the answer to, and being wary of those who claim to know it remains the best reflex.
What seems solid is that both camps are looking at the wrong thing. Those predicting a collapse in employment do not see it in the numbers. Those invoking history assume the pace will be the same, which is not a given.
The useful question is therefore not "will there be jobs", but "who will pay for the time of adaptation". That is a political question, not a technological one, and it arises now rather than in ten years. That it is being asked by an industry figure rather than by its critics is in itself information.