Revolution now

We are living through the birth of the Fourth Industrial Revolution: the rise of AI and cyber-physical systems. Genevieve Bell considers the first three waves to explore the possibilities of how to use them to build a better, brighter world.

I’d like to introduce you to a technology system built on the Barwon River near the Queensland/ New South Wales border. It’s in the middle of the town of Brewarrina. Today, it’s an archaeological site. Nearly 400 metres long, it comprises a series of stones, rock walls and channels. It’s the largest and oldest set of fish traps in Australia.

What this system did was create the capacity to trap fish going upstream or downstream, and to hold them in pens in cool running water – when the river was running both low and high. The reason for this trap here, on this river, was that it was a meeting place. It was a site where multiple Indigenous nations and families gathered, and where ceremonies, ritual and knowledge were built and exchanged. For that to happen, you needed to have food at scale, but it also meant people could gather again next year and for years to come.

The most remarkable thing about this system? Its age. For thousands and thousands of years, people used, adapted and modified this system, and even today, its custodians, the Ngemba people, still fish from the rock walls. Imagine building a technical system that would last for millennia – most of us are lucky if we build a system that lasts for 10. The Brewarrina fish traps (enshrined on the Australian National Heritage list in 2005) also reveal a deep understanding of technologies – in this case lithics – and an understanding of the ecosystem, hydrology, fish biology and an environment that was changing over this period of time.

Those three pieces – technical, cultural and ecological – are incredibly significant. As we think about building technical systems, we need to build them with multiple pieces in mind – not just the technology, but the ecological piece, and the human piece. It’s a useful way of framing how current technical systems should and could unfold.

In 2016, the World Economic Forum (WEF) published a chart that crystallised a conversation that had been around for a while about the notion of a Fourth Industrial Revolution. The WEF gave that conversation form, structure and a back story. It made it part of a series of earlier waves of history and waves of economic and technical transformation. In doing that the WEF stabilised the context of that history and its consequences, and also made a fetish of the technical systems. And in some ways they made quite mysterious the consequences of those revolutionary transformations to human society, to culture. What’s also mysterious about this story is the tale of who made these transformations possible, and the kind of practices and practitioners that had to develop.

We’re on the cusp of a Fourth Industrial Revolution. In order to consider its possibilities, I want to revisit the previous ones with an eye to two questions: what did it take to get to scale in each one of those moments, and what did scale look like?  

I: The Age of Steam

 The First Industrial Revolution starts in many places. One of them is Cornwall, England, in 1712, atop a mine, with a man named Thomas Newcomen. Newcomen was an ironmonger and preacher who occasionally helped save miners from the bottom of flooded mines. It’s in that latter context that he helped come up with the idea of the atmospheric engine – the prototype of all steam engines to follow. He was an inventor and an innovator – he took a whole lot of other people’s ideas and built them into a single object. Two storeys high, it was loud, it consumed everything around it – water and coal – and it changed everything.

It took 20 years before 100 of these objects were in circulation, and nearly 100 years before the number reached 2000. But in that slow, steady ramp of the atmospheric engine, with James Watt’s  transformations of it into the steam engine, you see an object that moves from mines into factories and changes the way factory work is done. It changes the possibilities of how things can be built. It makes a series of complicated, cultural and practical transformations.

But in some ways, the most important moment of scale isn’t that first century but the second. It comes in 1829 at a place called Rainhill in England, when a locomotive called Rocket changes the way people think about steam engines. They go from being stationary to mobile. They go from being powerful to being fast, and in so doing unleash the possibility of creating networks of trains and railway systems. But to get to a railway system required more than just a locomotive. It required regulation – regulation that fixed train prices so that everyone had access to the technology; that managed safety, so that the trains didn’t hurt people; that changed the ideas about how time should be configured so the trains would run on time. It took ideas about timetables and all kinds of new practitioners – ticket takers and safety inspectors and civil engineers. Those train systems unfolded in Britain and the US, Australia, Japan and India. They transformed the way we thought about time, about distance and about speed.

The irony of this is that this First Industrial Revolution was nearly two centuries in the making. Getting to scale in this instance took time, and it required all manner of regulations and social actors and practice to be accomplished.

II: The Age of Electricity

The Second Industrial Revolution is anchored on electricity rather than steam, but it includes many of the same challenges.

The first displays of electricity were at the Crystal Palace in England in the 1850s; as a spectacle at World Fairs; at the battle of light bulbs – the International Exposition of Electricity – in Paris in 1881. Even the Great White Way in New York City, where people went to see electricity, was all about persuading them they needed this new infrastructure. People already had ways of lighting their homes and powering things – they had steam, they had gas. Why did they need electricity? Part of the challenge for the Second Industrial Revolution was compelling people to imagine they should upgrade.

It also required whole new systems. It wasn’t just enough to make light and make power. Someone had to generate the power. There had to be an electrical grid. It required an incredible argument about the best way to configure that grid – AC or DC? How would you imagine all the work needed to upgrade and retrofit buildings? What new appliances and experiences would need to be created? It took individual actors in different countries and an entire apparatus to get to scale.

Thomas Edison, in the US, not only got lightbulbs down to a reasonable price, he also built the first electrical power plants in order to generate the electricity to make the lights work. In Britain, getting electricity to scale required civic and civil organisations. It required people like Caroline Haslett and the Electrical Association for Women, which helped empower British housewives and homemakers to troubleshoot their own appliances and keep things running. It required a whole range of conversations with businesses. And all of that, like the First Industrial Revolution, took time.

From the first experiments with electricity to the first electrification of power into manufacturing facilities took nearly 50 years. And the consequences of changing the way to power mass production was that we also had to build new kinds of companies, new kinds of business models, new kinds of corporations and even new kinds of practitioners. So – much like the First Industrial Revolution – getting to scale was a complicated puzzle..

III: The Age of Computers

Many of us remember the origins of the Third Industrial Revolution. It’s about automation and digitisation, and the technology required to get to scale here is the computer.

The earliest computers appeared in the 1940s, but you really don’t hit ubiquity of the object until the 1960s. What takes computers to scale in this instance isn’t just the availability of computers. It’s the creation of two things: programming languages that let people talk to computers and let them do things beyond the obvious, and the invention of a curriculum for computer science to help bring a whole lot of people into the conversation.

Starting in the early 1960s, there were a number of people all over the US using computers in their research and professional jobs. And starting in the mid-1960s, a whole collection of different people at American universities started to imagine what it would be like to teach people to use these computers, so that they weren’t just branded objects sitting inside companies, but could become platforms upon which many things could happen.

By 1968, an initial curriculum has been created to teach the subject called computer science (until then it was just “computing”). There’s now a curriculum. That curriculum does two things. It creates the possibility that a whole lot of people can have a shared experience and a shared vocabulary. It also creates the possibility that the way you can think about computing is no longer tied to the objects of now, but creates the possibility of other objects to come into existence.

So what it took to get a piece of technology to scale in the 20th century involved commercial enterprises, universities, the creation of a vocabulary and a set of questions that let many more people engage with those machines and do things with them that were unimaginable. It also involved the continuing evolution of the technical system and concepts like Moore’s law and companies like Intel and AMD [Advanced Micro Devices] continuing to revolutionise and innovate on those technical platforms. It involved the creation of a lot of other objects that took on computing power and made them meaningful. Again, it took time, and it took social actors. It took an extraordinary number of practitioners. And it’s still – much like the first two industrial revolutions – not entirely done yet.

IV: The Age of AI

That gets us to the Fourth Industrial Revolution and the notion of AI.

The first time the term “artificial intelligence” was coined was in 1956, in Dartmouth at the Summer Research Project on Artificial Intelligence. But really, we’d been talking and thinking about machines long before that – at the Macy Conferences in New York City in the 1940s and 1950s and with Alan Turing at Bletchley Park and Manchester in the United Kingdom. There’d been many conversations about what it might mean to make machines think like humans.

 When I think about AI, I return to its original articulation in 1956: to be able to so precisely describe intelligence that a machine can be made to simulate it. A machine that will understand human language, a machine that would understand abstractions and concepts, a machine that would be able to learn for itself, and a machine that would be able to ultimately do things that humans could do. AI requires data and machine learning and algorithms and sensors and some form of logic that holds all that together. But AI is not synonymous with machine learning, it’s more than that.

And when we talk about the Fourth Industrial Revolution we’re talking about more than just AI. We’re talking about AI as part of a system that is going to have to expand to include the systems in which AI finds itself: the Internet of Things, fast and rapid telecommunications protocols – all of it also built on the second and third waves of the Industrial Revolution of the electrical grid and the computing platform. It’s an industrial revolution that’s in the beginning stages. The technology is 60 years old, but what it will mean to go successfully to scale is going to take a whole lot of complexity and raise a whole lot of really quite complicated questions.

The interesting thing about scale is it’s often only visible when it breaks. The arc of the beginning of the 21st century has been stories that we tell about technologies that are reliant on things that we mistakenly thought were stable, like the massive power outages in California in 2020 and the damage to the electrical grid and telecommunications during Australia’s bushfires last summer. And not just technical systems: civic and civil society, ideas about democracy – all of these fields differently brittle.

Part of what has made the Fourth Industrial Revolution possible is the abundance of data, and data is often invisible until it’s played back to us, where we suddenly say, Hey wait, how do you know that? Scale is truly the logic that runs through the Industrial Revolutions, along with ideas about efficiencies and productivity.

I suggest that this fourth wave is slightly different from the waves that come before it. We have that history. It doesn’t just inform how we think – it can help us ask better questions about the future that is coming. I’m concerned about what it means to have safe, responsible and sustainable cyber-physical systems, not just productive and efficient ones. What would it mean to say that the systems we are building shouldn’t just go to scale, but that they should go to scale in a way that feels like something we can live with?

When we talk about the First Industrial Revolution we could also talk about the Luddites; we could talk about pollution; we could talk about the consequences of chopping down so many trees to feed those inexhaustible trains. If we talk about the Second Industrial Revolution and electricity we could also talk about what it means to have mass production and what that means to ideas about labour and to factory conditions. If we talk about the Third Industrial Revolution and computing we can talk about such things as privacy and trust and security and risk, and all of the energy it takes to make those systems functional.

 Previous waves were accompanied by aggressive boosters with utopian stories and equally aggressive people with dystopian narratives. Stories about demonic trains and runaway trains are rife through British newspapers and magazines in the 1800s. In the US there were fears that if homes had electricity, women and children would be vulnerable because they’d be more easily seen. When computers were entering the mainstream many voices spoke out against automation. A 1960s manifesto called The Triple Revolution critiques the appearance of automation, social revolution and defence funding, and asks, does this build a society that we want to live in?

We should be paying attention to this history when we’re imagining what it means to build the fourth revolution in a manner that we can all live with. Lurking in all the stories about the earlier industrial revolutions are stories about the people who made it possible. Not just the business owners and the inventors but the practitioners and the workers who made those systems go to scale safely – engineers, electrical engineers and computer scientists. Who will be the practitioners for that fourth wave?

 I’m convinced that we need something different for this fourth wave, which demands a different set of conversations and a different set of skills. I think what it actually requires is a new branch of engineering: which is what we’ve been doing here at the 3A Institute (3Ai). (It always helps to have spent 20 years in Silicon Valley; it helps you square your shoulders and say, Yep, I’m establishing a new branch of engineering.)

The good news in all of this is that establishing branches of engineering isn’t trivial. It’s also not new. The first engineering schools were established in Paris with the École Polytechnique, at Stanford with computer science, and at the Wharton School of Business. So there are lots of ways of thinking about this – but we decided we should start with some questions. We think there are six big ideas that help you think about how to build those systems, how to regulate those systems, how to decommission those systems and how to operate safely with those systems.

First question is: is a cyber-physical system really autonomous, and if so what does that mean? How do you engineer it, how do you secure it? How do you regulate it? How do you think about what that means from a policy point of view? Not all autonomy is built the same way. How do you think about whether those systems have agency, or put another way, what are the controls and limits on those systems and where do they sit? Are they inside the systems or are they outside the systems? Are they software, or are they hardware?

Second, who gets to activate them and under what circumstances, and how do you secure them? How do you evolve them over time? Because surely those systems will change over time – think back to the steam engines.

The third set of questions has to do with assurance. How do you know if the system is safe, reliable and trustworthy? How do you configure risk and liability? How do you think about privacy and trust, manageability, explicability, legislation, ethics and standards?

 Fourth, what will be the metrics we use to measure the cyber-physical systems’ performance? Much of the first three waves of the Industrial Revolution centred on ideas about efficiencies and productivities. We could also reasonably ask about safety and sustainability – how would we make those into indicators? What would the metrics be and who would get to look at them and think about them?

 Fifth, cyber-physical systems at scale also raise questions about interfaces and interactions. How are we going to engage with these systems? Will we engage with them? Will they engage with each other? What will that all look like, and what will it mean to have a whole series of technologies with which we are deeply familiar suddenly change their behaviour? If you put AI inside an elevator bank, the elevators behave differently. As humans, we stand there trying to work out why there are no buttons inside the lift compartment we’re in. How we build interfaces to those next generation technical systems is an open and interesting challenge.

And the last question, which is sometimes also the first question: what are the intentions of these systems? Why are they being built? What is the idea of the world that they are making and promulgating, and how might we critically interrogate that idea?

Backending those questions takes a little bit of cheeky and retro theory. I don’t think you can talk about cyber-physical systems without dwelling on “systems” and “cyber”. For me, that means going back to the work that was done in the 1940s by Norbert Wiener at the Macy Conferences. For Wiener, cybernetics was the scientific study of controlling communications and animals in the machine. For his colleagues, these were conversations about imagining a world where the system would need, by necessity, to include technology, culture and ecology. But if you go back to those early conversations in the 1940s and those first attempts to imagine what cybernetics might be, it was more about the notion of sophisticated technical systems than computing. While we now understand cybernetics as being about computing, it wasn’t always that way.

The earliest conversations at the Macy Conferences were about how the brain worked and how systems worked. They were about what we dreamed and talked about. For Wiener and many of those early conversationalists, cybernetics was as much about the system and the feedback loop as it was about the technology upon which it later came to be built. In reaching back to cybernetics as one of the theoretical underpinnings of this new branch of engineering, what I hope we’re also doing is creating space to imagine different ways of thinking about technology.

Revolution evolution

One of the most remarkable things about the Brewarrina fish traps is that it’s a technical system hundreds of years in the making and millennia in the using. It’s a technical system that required concerted and continuous effort. It wasn’t a quick fix; it required generations of accumulated knowledge about how the environment worked, about hydrology and fish, and also an accumulated commitment to continuing to build, sustain and upgrade that system over time.

I think about the fact that it was possible to be cybernetic without computing. I wonder what lessons we can glean from that original system – ideas about sustainability, about systems that are decades or centuries in the making, about systems that endure, that are built explicitly to endure, and systems that are built to ensure the continuity of culture. These feel to me like the kind of systems that we might want to be investing in now, in the 21st century.

To use cybernetics as a tool, it’s also about being able to bring through those older, more enduring stories, and finding a way to theorise them, alongside and in partnership with contemporary technology. So as we imagine these cyber-physical systems, this fourth wave of the Industrial Revolution, it isn’t just about how we build them and think about their economic consequences. It’s also how we theorise them, and in theorising them, how we ensure that we are asking the right questions about those systems and insisting that certain pieces sit inside those systems. Talking about an AI-driven world and an AI-ready society – a cyber-physical system world – without also talking about the people piece, and the environmental piece, is a conversation only half had. In England in the 1800s there was Greenwich Mean Time, but not everyone used it. Instead, the mayor would stand in the town square, and when the Sun was directly overhead the town’s clock would be set to midday. You get a bit of variance between London and Glasgow, and a couple of minutes isn’t much in the grand scheme of things – but when you’re running a railway it’s the difference between a catastrophic accident or not. The railways created their own time zone, and put a railway clock on every station.

For 40 years there was local time and railway time, until an Act of Parliament in 1881 made railway time standard. When I think about AI, I wonder not just about what it is that people imagine they’re doing with it, but about other “unintended consequences” that will be necessary in order for those systems to function effectively.

 What is the world you imagine you are building, and what does your imagination include and not include? Currently 10% of the world’s energy budget is spent on server farms. That doesn’t sound like a conversation you’d want to be silent to, but it’s not one of the issues we’re talking about. As we start to build the next industrial revolution, as we start to build a new branch of engineering, it’s not just about how we take those questions on technical systems, but also that we insist over and over that people and the environment need to be part of it too.

 The thing that being in Silicon Valley taught me, acutely, is that whilst we may tell stories about lone inventors, we never actually invent alone. So I need help. If you’re crazy enough to stand up and say you’re going to build a new branch of engineering, you should never do it alone. I think this is a conversation that should involve many, many different voices, as many voices as we can possibly sustain, because the Fourth Industrial Revolution requires a different approach than the previous three.

 In 1950, in a prologue for a play he was putting on at MIT, Norbert Wiener wrote that in the world that was coming perhaps the poets would need to be engineers or the engineers would need to be poets. The Japanese roboticist Mori says that our greatest anxieties about a computational other are actually just a manifestation of our own anxieties about ourselves. Ultimately the way these systems will learn is governed by the rules that humans write, and so if there’s a thing to be fearful about in all of this, it’s us.

Thirty years from now I hope this discipline finally has a name. I hope that much like their forebears in electrical engineering and computer science and civil engineering, practitioners of this discipline are part and parcel of building a world that we can all inhabit – a world that feels slightly more sustainable and a lot safer and much more responsible. For me, success would look like building practitioners who shaped a world that was a better place for all of us.