Exponential: Order and Chaos in an Age of Accelerating Technology

by Azeem Azhar

A quick survey of New York Times articles from a century ago reveals that Americans were apprehensive about elevators, the telephone, the television and more.3

It’s tempting to assume that people always feel technological and social change is too fast. They thought so a century ago, and they think so again now. But the argument of this book is that we are indeed living through a time of unusually fast change – and this change is being brought about by sudden technological advances.

the argument of Exponential has two key strands. First, new technologies are being invented and scaled at an ever-faster pace, all while decreasing rapidly in price. If we were to plot the rise of these technologies on a graph, they would follow a curved, exponential line. Second, our institutions – from our political norms, to our systems of economic organisation, to the ways we forge relationships – are changing more slowly. If we plotted the adaptation of these institutions on a graph, they would follow a straight, incremental line. The result is what I call the ‘exponential gap’. The chasm between new forms of technology – along with the fresh approaches to business, work, politics and civil society they bring about – and the corporations, employees, politics and wider social norms that get left behind.

An exponential gap emerges – between our existing rules around market power, monopoly, competition and tax, and the newly enormous companies that dominate markets.

In 1958, Fairchild Semiconductor sold 100 transistors to IBM for $150 apiece.5 By the 1960s, the price had fallen to $8 or so per transistor. By 1972, the year of my birth, the average cost of a transistor had fallen to 15 cents,6 and the semiconductor industry was churning out between 100 billion and 1 trillion transistors a year. By 2014, humanity produced 250 billion billion transistors annually: 25 times the number of stars in the Milky Way. Each second, the world’s ‘fabs’ – the specialised factories that turn out transistors – spewed out 8 trillion transistors.7 The cost of a transistor had dropped to a few billionths of a dollar.

Think of the first electronic computer, executing Alan Turing’s codebreaking algorithms in Bletchley Park in 1945. A decade later, there were still only 264 computers in the world, many costing tens of thousands of dollars a month to rent.8 Six decades on, there are more than 5 billion computers in use – including smartphones, the supercomputers in our pockets. Our kitchen cupboards, storage boxes and attics are littered with computing devices – at only a few years old, already too outdated for any modern application.

In 1974, Bill Gates said he envisaged a ‘computer on every desk and in every home’. At that time, there were fewer than 500,000 computers, of any type, worldwide. By the turn of the millennium, the number of computers exceeded 500 million – still fewer than one device per European or American home. Yet, within a couple of decades, a typical family in the West had half a dozen computers at home – between smartphones, the family computer, a modern TV, and a smart speaker like an Amazon Alexa. A gadget-savvy household could easily break into the double digits.

It took 11 years for social media to reach 7 in 10 Americans, at a time when the average lifespan of those alive when it was introduced exceeded 77 years. So social media took 14 per cent of a lifespan to achieve saturation. The comparative metric for electricity was 62 per cent of the average lifespan. Benchmarked against average lifespan at time of introduction, smartphones diffused 12.5 times faster than the original telephone.

After Facebook launched in February 2004, it became one of the fastest-growing products in world history, amassing its first million users in just 15 months.

Take TikTok, a social network for funny videos. It went from an unheard-of service to the most downloaded app in the world in a matter of months.

This increasing speed is the legacy of Moore’s Law. The hardware that underpins digital technology lends itself to continual increases in power and continual decreases in price. As chips develop at exponential rates – 50 per cent or more, compounding every year for many years – they give access to unimaginable computing power, for trivial sums of money. That hyperdeflation creates ever-greater possibilities: new products that can, in turn, spread faster through our economies. The overall process is one of constant acceleration.

For example, scientists hope quantum computers will allow us to create nitrogen fertilizers without releasing scads of carbon dioxide into the atmosphere. This is all about modelling new kinds of molecules, to be used as catalysts in the process of fertilizer production. Classical computers would take hundreds of thousands of years to model such molecules; a quantum computer would take about a day.

Between 1975 and 2019, photovoltaics dropped in price some 500 times – to under 23 cents per watt of power.3 The bulk of that change has come in the last decade. Even in 2010, it cost 30–40 cents (about 20–30 pence) to produce a kilowatt-hour of electricity using solar panels, making it 10–20 times more expensive than fossil fuels.4 But the cost of solar power has been declining at exponential rates – about 40 per cent per annum for large commercial contracts.

By October 2020, the cost of generating electricity from large-scale solar power, as well as from wind, had dropped below the cheapest form of fossil fuel product: combined-cycle gas generation.

In the decade up to 2019, the price of electricity generated by solar power had declined by 89 per cent.

Exponentiality has become widespread in four key domains of technology, which between them form the bedrock of the global economy. Computing, of course, is one. But so too are the domains of energy, biology and manufacturing.

Then, after a few years of dominance by Illumina, the Shenzhen-based company BGI announced in March 2020 it was capable of sequencing a full genome for only $100 – representing a million-fold improvement in less than 20 years.10

That equates to a halving in price every year for two decades. It puts Moore’s Law to shame. If the price-performance improvements in genome sequencing had followed the same curve as Moore’s predictions for microchips, at the time of writing you could expect to pay more than $100,000 per sequence. In fact, genome sequencing has fallen 1,000 times further than Moore’s Law would have forecast.

And, as ever, lower prices mean higher usage. In 1999, we had sequenced one messy genome. By 2015, humanity was sequencing more than 200,000 genomes a year.11 One research group estimates that by 2025 as m...

Since the first 3D printers were developed by Charles Hull in the mid-1980s, additive manufacturing has improved dramatically. The process has become faster, more precise and more versatile – today, 3D printers can work with materials including plastics, steels, ceramics and even human proteins.

Researchers estimate that most additive manufacturing methods are developing at a pace of between 16.7 per cent and 37.6 per cent every year, with the average rate falling in the high thirties.

Terry Wohlers, an analyst of the additive manufacturing sector, tells me that the 3D printing market grew 11 times in the decade to 2019 – a rate of 27 per cent per annum.

But the technologies in these four domains – computing, biology, energy and manufacturing – are special.

today’s technologies – general purpose or otherwise – are rolling out at a much faster rate than ever before. Much of the infrastructure, from cloud computing to smartphones, has already been deployed. And that means that the transformation Perez alludes to may come faster than in any previous era. Our age is defined by the cascading of technologies: one new technology leads rapidly on to the next, and on to the next.

This is the power of GPTs. Their effects spread inexorably from area to area, rippling across all aspects of our daily lives. And the GPTs of the Exponential Age are only just beginning to emerge. We have so far witnessed the rise of personal computers, the internet, smartphones; we haven’t yet experienced the effects of extremely cheap power, bioengineering, 3D printing – and many other technologies too nascent to predict.

His theory was that for every doubling in units produced, costs would fall by a constant percentage. The exact nature of the decline would depend on the engineering in question. In the case of the aircraft Wright studied, it was a 15 per cent improvement for every doubling of production. This 15 per cent improvement is known as the ‘learning rate’.

In Wright’s analysis, the reason was simple. As engineers build a product, they come to understand what it takes to build it better. They figure out a more elegant way to connect two components. Or they combine a set of different elements into a single component. Workers figure out shortcuts that make them more efficient. In other words, they learn by doing. As engineers finesse the process, a small innovation here and another innovation there drives rapid increases in efficiency.

This means Wright’s Law has an edge over Moore’s Law. Both describe the way the cost of technology diminishes exponentially. But Moore’s Law simply describes performance improvements over time. There are scenarios it can’t account for – those striking microchip-factory workers, for example. Wright’s Law, meanwhile, connects progress to the quantities produced. Say that for every doubling of production, the unit cost of a gadget drops 20 per cent. If production doubles every two years, costs will drop by 20 per cent every two years. If production doubles every year, costs will drop by 20 per cent every year. Wright’s Law holds true even during that mythical strike: if production stops, the reduction in cost stops.

But the greater cause of the newfound power of Wright’s Law lies in economics. Previously, the S-curve of demand tapered off when a market reached saturation. Today, that point of market saturation is much more distant – because global markets are so much larger. And this means that the process explained by Wright’s Law can continue for much longer, and the exponential gains can continue to mount up.

As we’ll see time and again in this book, the rise of a global market for products is one of the great changes of the last 50 years.

Bigger markets mean more demand; more demand means more efficient production; more efficient production means cheaper goods; cheaper goods mean bigger markets. The cycle has an inherently exponential logic. And so Wright’s Law describes how technological progress takes on its own momentum: the more we make of something, the more demand there is, and so the more we make.

Standardisation of this kind makes the world more efficient – and allows different innovations to combine. When technologies adhere to standard forms, they can be applied to a wider range of industries. Standard technologies become like Lego blocks that can be selected and assembled into an eclectic array of services. This combination and recombination catalyses further innovation.

In software, standardisation has become commonplace. It has given rise to a componentisation: many frequent tasks we ask of software are now available as easy-to-access lumps of code. A modern software developer may spend as much time sticking together such standard components as they do writing something new.

All this points to a shift in the nature of technological development. Modern technologies are more likely to combine and recombine. And this process of combination leads to the emergence of new innovations.

These preprint servers are so powerful because they remove the boundaries to academic research. They let ordinary people access cutting-edge ideas for free. And that widens the groups who can participate in the scientific process.

It is astonishing that an idea as powerful as generative adversarial networks can travel from a non-peer-reviewed white paper to a high-school project a continent away, all in less than five years.

Huge behemoths like Apple hold fewer than ten days of stock, or ‘inventory outstanding’. Millions of dollars of products that the company knows it will sell in two weeks’ time have not yet been produced. It can rely on the digital network of the internet to take the order and coordinate it, and a physical network of trucks and ships to meet all foreseeable demand.

When Apple launched its first iPhone on 29 June 2007, the device was available for sale in a single store in San Francisco. By the time the eleventh version of the phone was launched on 20 September 2019, it was delivered to eager shoppers in hundreds of different cities in more than 30 countries around the globe.

Exponential technologies are being driven by three mutually reinforcing factors – the power of learning by doing, the increasing interaction and combination of new technologies, and the emergence of new networks of information and trade.

In the UK, the basic rate of tax was 33 per cent by the latter half of the 1970s. And with high taxes came hands-on government: nationalised industries, heavy regulation and interventionist industrial policy. But Friedman’s acolytes advocated a different approach. By rolling back regulations and cutting taxes, governments could unleash the power of the market – and bring the return of high growth and manageable inflation. This market-friendly ideology found its touchstone with Friedman’s famous doctrine, which held that the social responsibility of corporations and the business sector was to increase profits, and not much else.

Free-market economics helped drive globalisation. And globalisation helped kindle the Exponential Age. All this means we can trace the beginnings of an era of exponential technology not just to the discovery of Moore’s Law or the creation of the seminal Intel 4004 processor in 1971. Just as important was the emergence of a new political orthodoxy. From the late 1970s onwards, free-market capitalism would unleash the power of exponentiality.

At the heart of Amazon’s success is an annual research and development budget that reached a staggering $36 billion in 2019, and which is used to develop everything from robots to smart home assistants. This sum leaves other companies – and many governments – behind. It is not far off the UK government’s annual budget for research and development.1 The entire US government’s federal R&D budget for 2018 was only $134 billion.

Perhaps more remarkable is the rate at which Amazon grew this budget. Ten years earlier, Amazon’s research budget was $1.2 billion. Over the course of the next decade, the firm increased its annual R&D budget by about 44 per cent every year. As the 2010s went on, Amazon doubled down on its investments in research. In the words of Werner Vogels, the firm’s chief technology officer, if they stopped innovating they ‘would be out of business in 10–15 years

This kind of linear thinking, rooted in the assumption that change takes decades and not months, may have worked in the past – but not anymore. Amazon understood the nature of the Exponential Age. The pace of change was accelerating; the companies that could harness the technologies of the new era would take off. And those that couldn’t keep up would be undone at remarkable speed.

Such a divergence between the old and the new is one example of what I call the ‘exponential gap’. On the one hand, there are technologies that develop at an exponential pace – and the companies, institutions and communities that adapt to or harness them. On the other, there are the ideas and norms of the old world. The companies, institutions and communities that can only adapt at an incremental pace. These get left behind – and fast.

The emergence of this gap is a consequence of exponential technology. Until the early 2010s, most companies assumed the cost of their inputs would remain pretty similar from year to year, perhaps with a nudge for inflation. The raw materials might fluctuate based on commodity markets; but their planning processes, institutionalised in management orthodoxy, could manage such volatility. But in the Exponential Age, one primary input for a company is its ability to process information. One of the main costs to process that data is computa...

In general, if an organisation needs to do something that uses computation, and that task is too expensive today, it probably won’t be too expensive in a couple of years. For companies, this realisation has deep significance. Firms that figured out that the effective price of computation was declining, even if the notional price of what they were buying was staying the same (or even rising), could plan, practise and experiment with the near future in mind. Even if those futuristic activities were expensive now, they would become affordable soon enough. Organisations that understood this deflation, and planned for it, became well-positioned to take advantage of the Exponential Age.

The bosses at Tesla understood that the prices of electric vehicles might decline on an exponential curve, and launched the electric vehicle revolution. The founders of Impossible Foods understood how the expensive process of precision fermentation (which involves genetically modified microorganisms) would get cheaper and cheaper. Executives at space companies like Spire and Planet Labs understood this process would drive down the cost of putting satellites in orbit. Companies that didn’t adapt to exponential technology shifts, like much of the newspaper publishing industry, didn’t stand a chance.

In those early days, exponential change is distinctly boring, and most people and organisations ignore it. At this point in the curve, the industry producing an exponential technology looks exciting to those in it, but like a backwater to everyone else. But at some point, the line of exponential change crosses that of linear change. Soon it reaches an inflection point. That shift in gear, which is both sudden and subtle, is hard to fathom.

Exponential processes are counterintuitive. And we struggle to grasp them. Thomas Malthus, the eighteenth-century political economist, first articulated the problem. According to Malthus, the human population tends to grow exponentially – but we won’t realise the power of that exponential growth until too late. Eventually, human needs will outgrow our ability to produce food, bringing famine and pestilence. Malthus’s dire predictions did not come to pass, thanks to the extraordinary increases in productivity brought about by the industrial revolution. But his basic insight – that we continually underestimate exponential processes – remains pertinent.