Daniel Susskind, an economist at King's College London, has written a short book for parents who no longer know what to tell their children about work. What Should My Children Do? argues that the right response to AI in education is absorption rather than resistance, and it rests that case on an unlikely foundation: a 1979 British government report about pocket calculators. Susskind, a former policy adviser in the UK Prime Minister's office who also researches AI ethics at Oxford University's Institute for Ethics in AI, is explicit about the parallel. He sees himself as a sort of Cockcroft of the 2020s.
Wilfred Cockcroft was a nearly forgotten maths lecturer at Hull University when the government asked him to report on the crisis in teaching the subject to children. His document ranged widely, and it dealt at length with the consumer technology of the moment. Parents and teachers feared the calculator would erode arithmetic, one of the traditional three Rs alongside reading and writing, and produce generations with weakened faculties who reached into a pocket rather than work a hard sum out. Cockcroft went the other way. Adopt the calculator instead of fearing it: teach children to count with the device and without it, then test both skills.
Susskind's reading of that episode is that Cockcroft did not pretend the oncoming wave of technological progress could be walked around.
The analogy has limits, and the book does not hide them. Even allowing for the AI industry's self-regard, what it has already shipped has changed enough that the calculator looks roughly as consequential as a corkscrew. But the scale of the present shift is only partly unprecedented, and the historical frame does make it easier to look at calmly. That is the book's whole method: find the earlier moment when adults panicked about a tool, and read the panic back against the record.
The title is the tell. It captures the specific technological anxiety of parents trying to picture a future under extreme uncertainty. Adults used to pass on accumulated knowledge with a rough sense of how the next generation would use it. That continuity has gone, Susskind writes: you cannot be confident the advice you gave your child last year will still be useful next year.
Susskind is raising small children himself, and the future he is trying to picture is one of artificial general intelligence — AGI that surpasses humans at all economically meaningful tasks. What people currently deal with through ChatGPT, Claude and similar systems is narrow AI, good at individual tasks. In an AGI world it is not just whole professions that change but the meaning of work itself. Read the book and that scenario feels closer than it did before you opened it, which is why the author pushes politicians, parents and teachers to begin adapting now.
His sharpest warning is against trying to future-proof when the horizon is this foggy, and his example is a good one. David Cameron's government hurried to make learning to code a priority for children a few years before coding became one of the things AI does best. The lesson Susskind draws is that specific bets on specific skills age badly. Schools should instead teach children to acquire knowledge both with AI and without it, while shoring up the basic skills that must not be lost from view.
What follows is a picture of personalised learning. Teachers trained to work with AI use it as an assistant and a mentor rather than trying to instruct thirty children as one undifferentiated group. Graduates entering the office industries most exposed to change use AI to shed the routine work — reviewing legal documents in a basement is his image — and build professional judgement earlier, on AI-generated simulations. Mid-career people are told what not to do: fall back on inherited practice, convince themselves their profession is safe, forget which original problem their profession exists to solve, or refuse to rethink how that problem gets solved now. The instruction is to remember the underlying task and redesign its execution around AI.
The Cameron anecdote is the best thing in the book, and I think it quietly undermines the prescription that follows it. Cockcroft could say "test both skills" because there was an examination system capable of enforcing the second half of that sentence. Susskind's version — learn with AI and without it — has no equivalent enforcement anywhere in sight, and a piece of advice that is correct in every possible future is also a piece of advice that rules nothing out. The book is admirably honest that the horizon is unreadable; it is less honest about what that honesty costs its own recommendations.
What the small book does not find room for is the more immediate ethics of using these systems: their dependence on stolen creative material, the environmental damage, and the questionable personal character of many AI company executives. That omission is not a side issue in a book addressed to parents. A parent asking what their children should do is also asking whose products they will be working inside and whose firms they will be working for, and those are exactly the questions the book routes around on its way to the calculator.
Cockcroft's report worked because everyone already agreed where the floor under arithmetic sat, and an exam could be built on it. Nobody has said where the floor sits under AI. Susskind asks parents and schools to hold one anyway, without naming it.