Why we must use AI in schools

It is beholden on us to prepare children for world as it will be. It's time to take The Sniff Test.

The City of Startups

London is a hotbed of startup activity. Startup Genome ranks it as the clear leader among ecosystems in Europe and tied with New York in second worldwide, behind Silicon Valley. Tech talent, investment and networks are concentrated in London.

Last week I attended a gathering of crypto enthusiasts and chatted with the owner of an incubator for AI and blockchain entrepreneurs. Each one is guided through their first 12 months and given help raising money. There are 300 applicants for the eight places in every cohort.

The owner is visiting senior schools, looking for one that teaches children about AI. He has yet to allow his daughters mobile phones and insists on a gradual introduction to technology. Every school he sees says it bans AI.

There are three things that could be taught about AI. The first is how it works, the second is its enormous potential and accompanying risks, and the third is how to use it. Before we get to each, it’s useful to consider what education is for.

The Purpose of Education

Nick Gibb, the Schools Minister in the Cameron government, gave a speech in 2015 about the purpose of education. He said it was the engine of the economy, the foundation of culture and essential preparation for adult life. This is the perspective of the state. For an individual, education is how to get on in life, engage your brain and guide you to what you will do when grown up.

Education has always been a route to prosperity. This requires literacy, numeracy and for long periods, mastery of a foreign language such as Latin or French. There is a strong social aspect to schooling but that comes regardless of what is taught. Reading, writing and arithmetic are foundational.

What it means to be literate changes. I don’t write or spell like Shakespeare and language evolves as a vehicle of culture. If literacy requires an ability to write concise emails, sharp marketing copy and converse with a large language model, why would we not teach these in school?

Gibb argued that a basic mastery in mathematics is essential to success in our modern economy. Ignoring the oxymoron, is this the case? Do people struggle with debt because they cannot calculate an annual percentage yield, or is it because they need to borrow? Few jobs require maths beyond what can be performed on a calculator.

Yet career progression does require numeracy and what this means evolves just as literacy does. The more senior you are the more you sell, and arithmetic matters in contract negotiations and budget reconciliations. A knowledge of statistics allows you to interpret data and is important even if you write for a living, because you must analyse your audience.

When I was at school, I was shown the steps to solve algebra puzzles and I learned and applied them. I don’t remember being told why I might want to know them. Now children can be shown those same steps by an AI model that also explains their purpose. The more dependent on technology we become, the more an understanding of why rather than how is the purpose of education.

A good memory is important for academic success. Children are turned off school because a subject is boring, or because they cannot store and recall facts. Now they have access to the best memory on the planet. If they are taught how to ask a precise question, they will extract any information they need. This is of huge economic and cultural value.

Knowing stuff that others know is how culture is created and people thirst for knowledge. As a child I knew the grounds and nicknames of all 92 clubs in the Football League. Had I been interested in what I was taught about Russian composers, I would have hung with a different crowd. Using AI to interrogate the internet, I can now get up to speed far faster than before and widen my social network. Mixing with people is the number one way to advance in society.

Access to knowledge opens up opportunities for children in choosing careers. From the age of five I wanted to be a stockbroker, but it took my whole education to acquire the skills to gain an entry level position in the profession. Michala chose to study psychology and her university required an understanding of chemistry. Good schooling provides the building blocks for us to find who we want to be, but it is available to too few.

Despite being academic, my most vivid school memories are of sport. Michala’s are of music. Kids want to be footballers and performers, but few make it and education provides a practical path to an alternative and comfortable life. At the same time, education uncovers those who will excel in different disciplines and if we all have AI at our disposal, how will we find those people?

The Creation of Elites

Let’s define the elite as the people who change things. Change is progress, by which I mean moving through progressively higher levels of problems. The faster we acquire foundational knowledge, the quicker we move on to solving new issues. This is what Newton meant when he said he stood on the shoulders of giants.

Elites are different to experts. Being a subject matter expert is not the same as advancing knowledge in that subject. Often it is the opposite, as with AI ethics experts who are philosophy dons with no background in technology, and are environmentalists in disguise. My argument for the end of the age of experts, rests on a better class of leader able to distinguish between those who favour progress over the status quo.

The pace of breakthroughs is accelerating as illustrated by three stories from Nature this week. AI enabled researchers to uncover 70,500 previously unknown viruses, which will provide a better understanding of disease and how the environment works. An AI tongue can distinguish between Coke and Pepsi, offering a different way to use existing materials to solve old problems. And ChatGPT scores better than doctors at empathy when breaking bad news to patients.

Education serves the purpose of surfacing the changemakers. These are the people who fall in love with their subject or their calling, and push its boundaries. These people emerge faster in the age of AI.

How to Teach AI

For every technology there are the few who build and the many who use. Most of us have no need to know how AI works and the matrix mathematics may be beyond us, even if taught by AI. What matters is an understanding of the potential of technology.

I am currently engaged writing copy for AI services. There is a marked distinction between the use of AI for data science and university research, and the consumer facing large language models. I’d wager most people have minimal understanding of the breadth of AI and tend to focus on stories of it surfacing nonsense from the internet and being used by kids to write essays.

Why shouldn’t they write with AI? The most important part of an education is how to use the available tools. For centuries we’ve judged intelligence with academic tests and machine learning is no different. By training AI to pass exams we’ve made the school curriculum redundant.

This presents a wonderful opportunity. Rather than have children rehash an argument about the causes of the First World War, let them generate arguments from two different viewpoints and debate them. Have the model produce a C-grade and an A-grade essay and then discuss the difference in class. Let’s have children spending more time thinking why and less memorising of what.

Let’s teach our students to validate the conclusions of research and build their own predictive models. A dry maths subject such as trigonometry is essential in architecture, navigation and computer graphics. Creating virtual worlds and exploring them will show children how to put the concepts into practice.

Education will become entertaining. Celebrity avatars will front models providing personal tuition, teaching with infinite patience and no prejudice. Thomas Jefferson can tell you about the Constitutional Convention and Einstein about the Theory of Relativity. You can ask them questions without fear of sounding stupid or being beaten up for being a swot. There will be an emphasis on finding and solving problems.

Instead we see fit to ban these tools and to overwhelm people with speculative nonsense about the dangers of technology. Europe is the world leader in this and being left behind. The desire to reframe everything to one political view, reshapes the AI debate into one about energy resources.

Azeem Azhar teamed up with BCG Henderson to model hyperscaling of AI. They conclude that GenAI inference, a fancy term for running models which is by far the most consumptive part of the process, would absorb at most 34% of available power supply. Technology companies are funding green energy projects to power data centres, and AI models enable optimisation of power meaning they will pay for themselves.

AI is here to stay and ignoring it is a dereliction of duty. Change requires strong leaders to persuade teachers unions and invigorate an inert civil service. As a result, my new friend will not find the right school for his kids. But if education is about preparing our children for the adult world, it’s beholden on us to teach them what will be in that world.