Daniel Susskind has spent 15 years studying how artificial intelligence affects work and society. As a father of three young children, he says the topic has become personal.
He recently asked ChatGPT to write a story in the style of Dr Seuss for his eight-year-old daughter. It took only a few seconds to produce.
The story wasn't perfect. Some lines didn't make sense, and a few American phrases had crept in. But his daughter enjoyed it, and the family spent the evening tweaking the prompts together.
A few weeks later, Susskind's family listened to a podcast called History's Not Boring during a car trip. The show features two children narrating short history episodes.
The family assumed the children were real. They later learned the entire podcast, including the young voices, was generated by AI.
Coding Skills and the "Future-Proofing" Problem
In 2013, the UK government announced that all schoolchildren would learn to code. The plan was meant to prepare students for future jobs.
Susskind points out that AI systems are now especially good at writing code. In January 2026, Anthropic said 90% of the code for its Claude Code assistant was written by AI rather than humans.
He argues this shows the flaw in trying to "future-proof" children by teaching specific skills. No one can predict which skills will still be valuable years from now.
Instead, he says schools should accept this uncertainty and focus on skills that remain useful no matter how technology changes.
Getting Back to Basics
According to the OECD's Programme for International Student Assessment, literacy and numeracy scores among young people have declined since 2009. The same trend applies to adults.
Susskind says this makes basic skills more important, not less. He explains that AI models still make mistakes and sometimes generate false information with confidence.
Because of this, he argues that people need strong reading and math skills to check AI output rather than accept it blindly.
He points to a British mathematician, Wilfred Halliday Cockcroft, who studied UK math education in the 1970s. His 1982 report responded to worries that calculators would ruin basic math skills.
Cockcroft's solution was to divide math teaching into two parts: one section using calculators, and one without. Both were tested separately.
Susskind proposes applying the same model to AI in schools today. He calls it "teach both, test both."
Under this approach, students would learn to use AI tools in one part of a subject, and complete the other part without any AI assistance. Both sections would be assessed.
He says this method would let students benefit from AI's capabilities while still building the basic skills that support all other kinds of learning.
Susskind adds that experimentation is already happening in some classrooms, including projects that combine AI tools with subjects like literature and history.
He says these efforts show what is possible when AI is used to support learning rather than avoided altogether.
