From global to local: a new (old) model for the future of education
In a world increasingly mediated by a layer of artifice that is getting harder and harder to differentiate from reality, how can we create spaces that enable authentic living, working and learning?
There is much to be optimistic about as we watch the rise of AI’s intelligence, versatility and sheer utility. Applications like ChatGPT and Midjourney constantly surprise and delight us with their output, and we are becoming adept at how to work with them to get the results we desire. But with every positive turn there are worrying signs on the horizon, storm clouds gathering that are making us ask questions about the very nature of reality in a world where AI, in the words of Yuval Noah Harari, has ‘hacked the operating system of human civilisation’.
Deepfakes and voice cloning are two such developments. Criminals are now cloning voices of loved ones and making spoof calls, gaining access to passwords and other confidential information. In one disturbing recent story, a woman believed her child had been kidnapped. It turned out that the supposed kidnappers had cloned her daughter’s voice. Language, both spoken and written, is fundamental to our operating system, and AI can hack into both in increasingly sophisticated ways.
Visual symbols are also under attack. Deep fakes are on the rise and becoming increasingly difficult to distinguish from reality. For example, I created the below in less than 5 minutes within minimal skill by downloading an image mapping AI from Github:
So what can any of us do about this? It seems pointless believing we can shut the stable door. Pandora is not one person nor even a group of developers: Pandora is a vast human/digital intermeshing with no obvious off-switch. The notion of time ceases also to mean anything: Moore’s Law was predicated on human ability to improve IT processing power roughly by doubling its speed and halving its cost every two years. But AI is already showing this to be null and void: if a whole universe of knowledge could be created in a millisecond, then we could find AI reaching the singularity at any point, well before we knew what had happened.
If it all sounds rather bleak, I believe within this bleakness is actual hope. Because, if we can no longer trust the world as it is presented to us through the sophisticated mediations of AI, then our only recourse is to turn our gaze inwards, into our local communities, privileging direct human interface over digital communication and teaching children within newly formed communities of learning and mutual care.
And if this sounds utopian it actually isn’t. As this is what the world was like before the invention of the steam engine.
How children learned before steam
Before there was a need to industrialise education in order to create the workers we needed for the cities of the past, there were four main ways in which children learnt what they needed in order to take their place in society as they got older.
Observation and participation
First of all, children learnt through observation and participation. They would watch the adults (not only their parents) and older children within their village and gradually begin to participate in activities like hunting, gathering, farming, cooking, and crafting. This allowed them to gradually acquire the skills and knowledge they needed in order to be useful contributors to their community. This was as much about trial and error as it was about formal instruction: most working people would have had little time to sit down and patiently explain how things worked. When you were involved in acts of subsistence others had to work it out for themselves, learning fast so they could ensure they were useful. The idea of learning styles, special needs and so on was irrelevant: children would work out what they could do within their zone of ability but also within the context within which they grew up. The reason we have surnames in most of the world is often linked to our job: Smith is one of the most common names in the UK because so many families were blacksmiths, such was the importance of horses in pre-industrial Britain.
Mentorship and apprenticeship
As children grew, they might also learn through mentorship and apprenticeship, with older individuals taking on the role of mentor or teacher for the younger ones. This could take the form of a formal apprenticeship, for example into a smithy, or could be a less formal relationship. The role of the mentor was to guide, encourage and support, again not necessarily to formally teach.
Stories
Another key teaching tool was the story. Before book-learning and writing were important, much learning came through the oral tradition. Elders would share stories, myths and legends that conveyed important cultural knowledge, moral lessons, and explanations of how they believed the world worked. Before the enlightenment, the story was central to our understanding of the scientific world, and although we may now scoff at arcane explanations of the seasons being the creations of Persephone or Freya, there is a beauty to these stories which helped draw communities together.
Play
Finally, children learned through play. Children still naturally learn a great deal in this way: problem-solving, sharing, creativity and physical skills can all be trained in this way. Pre-industrialisation there was no need for drilling children in literacy and numeracy from the age of five: children could be themselves, enjoying being children until their natural curiosity took them away from ‘childish things’ towards the wider world and how they could be an active and worthwhile participant in it. Nowadays it almost seems anathema to some education systems to allow free play without some sort of overlaying structure imposed on it: even the UK’s early years foundation system, which encourages free play, still has limitations placed on it. There are some countries in Europe who don’t expect any sort of formal, full-time education until the age of seven (Finland, Latvia and Switzerland among them) and they have some of the most highly regarded systems in the world.
Moving away from the city structure
The reason this is worth referring to is that, rather than move us even further away from this pre-industrialised model, AI can actually draw us closer to it, whilst still preserving many of the benefits of living in the modern world. After all, we may have been closer as communities before steam, but we also had the plague, women burned alive for supposedly being witches, and massive infant mortality. So it’s not all doom and gloom living in the 21st Century, however much we might yearn for a past long gone.
The reason AI has this potential is because it removes much of the need for the city structure. We saw through Covid that the majority of workers could continue with their professions from their homes. Not all of course: businesses such as hospitality and live entertainment were decimated. However, the majority of knowledge workers were able to continue from home, and many found their productivity increasing as a result.
The reason cities were created was to move ‘units of productivity’ as close as possible to the source of production: in other words, to move workers closer to the factories. As the industrial revolution robbed many subsistence farmers of their livelihoods, so they moved to the cities in hope of a better life.
The statistics are remarkable: in the first fifty years of the industrial revolution the population of England and Wales nearly tripled, from 7.5 million in 1801 to 21 million in 1851. The majority of this growth was concentrated in cities, as the increasing proximity of people to one another caused a ripple effect, increasing the rate of growth exponentially. In 1801, about 20% of the population lived in urban centres of more than 10,000 people. By 1891, that figure was closer to 75%. London saw the largest growth, moving from about one million in 1801 to over 6.5 million by the end of the 19th century. It’s interesting to note that in the 120 or so years since, London has only grown by about another 50%. This shows the power of the steam engine in fundamentally changing how society was organised.
Whilst the steam engine led to the greatest economic progress of any invention in history, in many ways it was the worst possible thing that could have happened to local communities. The young escaped the villages leaving the elderly to die alone, children were raised in dirty, cramped and squalid conditions, crime rose exponentially, and disease was rife. In mid 19th Century Liverpool, life expectancy was a staggeringly low 15 years, against the 52 years in rural Rutland.
But with AI we have the opportunity to turn back the clock, to regain an understanding of the local and to rediscover the joys of looking after each other and focusing on what we can actually control. It is true that so much of our anxiety stems from seeing a world falling apart and believing we can do nothing about it. We see global warming, moral decay and warfare and it depresses us because we feel helpless. And whilst AI cannot yet solve these problems, there may come a time when it can. As Yuval Noah Harari has recently stated, this is the first time in 3 billion years that we have had a non-organic form of being in the world. Whilst this being may not yet have the advanced and nuanced intelligence of humans, it is accelerating at a pace far beyond the ability of humans to learn. Only a few years ago Open AI’s GPT had the strategic intellect of a 4-year-old. As of December 2022, that had increased to a 9-year-old. Now, with ChatGPT-4, it is able to strategise as well if not better than most intelligent adults. How long will it be before it can offer (and execute on) genuinely beneficial solutions to the environment, to disease and to world poverty?
What AI can do, more immediately however, is enable us to recreate genuine communities. This is because, for the first time, we can genuinely rewrite the book on how we educate our young.
From school to hub: creating new communities of living, working and learning
To date, technology has had little profound impact on education. The main reason is that we have had no mutually beneficial relationship with the devices that have powered our lives since the dawn of the personal computer back in the 1980s and the growth of the internet in the 1990s. They have served a purpose, been functional, got the job done often far better than we can. We would not have made the advancements in science (such as the discovery of the Higgs Boson) if computing power had not proven our hunches correct. Our knowledge of the universe was largely conjecture until massively powerful, computer-controlled telescopes like James Webb pointed their lenses into the far reaches of space, enabling us glimpses of life’s origin.
However, what we have not had, until now, is a relationship with a technology that enables both sides to progress. But this is what we now have in AI. The nature of AI’s wiring, to learn through reinforcement, means that it continually improves in its responses and understands us the more we interact with it. ChatGPT-4 is a good example of this. You can ‘prime’ the AI before you begin to work with it, for example by inputting text, asking it to sample that text for its style, and then reminding it as you progress to ensure it stays on track. Its relatively short term memory (what is known as model size) is something of a limitation at the moment, but that is improving all the time. Whereas the first ChatGPT model had a memory limit of around 3,000 words, GPT-4 has double that and is soon to have something like eight times that amount. Anthropic’s Claude large language model is already showing a word memory of around 75,000 words, which means you can input who novels and work with the AI to analyse or even write new storylines based on the style of the original author. We can imagine whole new works from the likes of Charles Dickens, or Karl Marx reflecting on the impact of his ideas on the 20th Century (and perhaps revising his Manifesto as a result). This malleability and adaptability is the key to its potentially massive impact on learning - and as a result, on the spaces within which this learning can occur.
To now, there has been a limit on the reach of any one individual teacher and their ability to both control, and teach, groups of children. As a result, schools have been stratified into year groups, subjects, and often ability levels. Children from different ages (and often different abilities) do not mix, the structures of the school supporting this: the lower school playground or special needs centre, for example. Now, there is a place for more therapeutic environments in a school: autism centres with their sensory rooms, calm lighting and soft surfaces being a good example of this. However, for most children, it makes little sense to separate them one from another, regardless of their age, stage or learning needs. This is not how we work in the world, so why do we perpetuate it in schools?
The reason is that we have needed a specialist early years teacher, or year 5 teacher, or A-Level English Literature teacher. And whilst I am not saying that the nature of specialism will die out completely, I think it will become increasingly less important the more bespoke AI mentors support children from a young age and can work with any subject to any degree of specialism. We have already seen ChatGPT-4 ace the Uniform Bar exam, considered one of the toughest placement exams in the world. And remember: the AI we have now is the worst it will ever be. The nature of reinforcement learning means that it is improving faster than any of its human developers thought possible.
Consider this for a moment, as it is a critical point. In fact, the entire education system rests on this one premise: schools are schools because there is a limit as to the number of age groups, number if children, or degree of specialism, one person can manage. That is it. We have classrooms because, by and large, adults find it challenging to control more than thirty children. We have graduates moving from their maths degrees into teaching IB maths because someone needs the specialism to take them through the course. But if AI can craft the perfect learning pathway and support groups of children through immersive, real world projects, and the ‘fan’ model I explored in an earlier article inverts into a ‘hub and spoke’ model, with coaches, peers and even external expertise supporting the child at the centre of learning process, then what happens to our school buildings?
In two words: they disappear.
The learning spaces of the future will not be housed inside schools. That form is increasingly less fit for purpose. Learning will become localised and decentralised, with community hubs supporting both learners and workers. After all, if the school can be decentralised then so can the place of work. If we believe that the foundations of a strong society stem from strong, well-supported people who are connected with one another and feel a sense of belonging, then it does not take much of a leap to see that, if we can keep families working and learning together as much as possible, that has to be good for society. So many of society’s ills come through what Michel Houlebeq called ‘atomisation’: a fracturing of society into individuals, floating around in a sea of their own anxieties, cut adrift from the flow of life in all its vicissitudes.
A localised learning hub will look towards its community. It will not be concerned with global politics, fake news, or mass media manipulation. It will ignore deep fakes and voice cloning because it will deal with the immediacy of human presence. If we can no longer trust the world as it is mediated through technology, then our only choice is to turn our lens away from that world and to focus on those around us. There is a simple beauty to this truth, and it is one that I believe is the only way forward. We can look on how much of the world will become divorced from reality in the next few years with horror, decrying a time before when we could pick up a newspaper and see photographs we knew were taken with a camera loaded with film, or watch a TV show and know that the actors were flesh and blood and that the words they spoke came from the creative mind of a brilliant writer. We can lie awake troubled by how to protect our children from the growing panic of voice cloned scams or the continual rise in online grooming. Or we can see it as an opportunity to move back to what I call techno-agrarian communities, digital villages if you like, small groups of likeminded people who live, work and learn together. I don’t necessarily mean that the world will move to a 1960s communal model, although for some that might be very attractive. But small, nurturing spaces where everyone knows everyone, where children learn from many different adults and older peers, and where there is return to mentorship and apprenticeship: is that really such a bleak vision of the future?
We can imagine one of these hubs. Each morning the whole family would arrive early and enter one of the quiet rooms to engage in yoga, meditation and other forms of grounding practice. From there a healthy breakfast could be taken in one of the organic cafes dotted around the hub. The children then move off into a separate, secured space for their learning activities for the morning, consulting with their AI ‘genie’ to remind themselves of what they need to be focusing on that day. Groups are not split by age, gender or ability: older and younger children work together, supporting and learning from each other. The adults would move into one of a number of different spaces: quiet spaces to work on their laptops, small meeting rooms for taking face to face or online meetings, or creator spaces for content creation and other entrepreneurial activities.
Lunchtimes could be spent back as a family or children could relax with their friends. There would be a degree of flexibility. Smart tracking would mean that there was no possibility of children wandering off unaccompanied, and besides, the mentorship programme would be so strong that often the younger and older children would hang out together like siblings. Children would not be glued to their screens during break times, unless they were engaging in e-sports or other gaming activities during certain set periods. They would be outside, in nature, playing. Being children.
Afternoons would be given to a programme of arts, music, drama, dance and sports. Every child would learn a musical instrument, would take up an artistic pursuit, and would be involved in some form of creative and athletic movement. The key here is that parents and other members of the community could be involved in this afternoon programme, not only as coaches but also as learners. Mother and daughter could be learning to play guitar together, or the whole family could be involved in a drama performance or learning how to grow vegetables.
This is not some utopian dream. This has to happen. The reason is simple: future schools won’t be places where we train children to become part of the knowledge economy, because the processing and actioning of knowledge, data, and language will be automated largely by AI. If AI has hacked our operating system, we need to consider how we can better operate within a new system. So much of what we currently teach is based on what children needed to know, understand and be able to do more than a century ago. And most of this is no longer valid. What we need instead to do is to teach children to be human, because that is the one thing AI will never be. It might be able to approximate human emotions and connect with humans in an incredibly accurate way, but it will never be truly authentic. AI has no limbic system; its brain did not grow over millions of years into the intricate, beautiful, complex and often flawed organism that is the human mind. As Manolis Kellis has rightly stated, it is like AI went straight for brain version 3.0 without any of the earlier iterations. This means that AI cannot connect with us on this messy, emotional, human level. Which means these are the skills we need to be nurturing in ourselves and our children. What an incredible opportunity we have now.