Data is not the new oil – it’s the new electricity
The most powerful technology yet invented by man has had virtually no effect on productivity and, by extension, the overall standard of living. Yet computing was never about the technology, but always about the information. And information, to quote Heinz von Foerster, is “order wrenched from disorder.” With AI algorithms now doing exactly that to the ever-increasing store-houses of digital data, computing technology might at last be on the cusp of fulfilling the destiny of any general purpose technology: to transform everything. The best way to help it do that, writes Dominic Hobson, is to place the ownership of data in the hands of the people that create every company that exists: consumers.
The most enduring mystery of the computer age is why information technology has had so little impact on our standard of living.
Despite reducing the marginal cost of production to zero, computing has – an upward blip in the United States in the late 1990s apart, and one almost entirely attributable to an improvement in logistics at Walmart – had next to no effect on productivity in the United States or the United Kingdom.
Why hasn’t computing raised productivity?
The pithy observation by the great pioneer of economic growth models, MIT economics professor Robert Solow, that “you can see the computer age everywhere but in the productivity statistics” is as true now as when he first made it in 1987.
In The Rise and Fall of American Growth, his magisterial study of the economy of the United States since 1870, Robert Gordon concludes that computers cannot compete with electricity, the internal combustion engine or even clean water and public health as a driver of economic growth.
Yet computing is, like electricity, a general purpose technology. Indeed, it is the most generally purposeful technology of all time.
Computing is free of any physical sub-strate and unequivocally universal – even the most inert rock formation encodes and processes information at the atomic level. Which is why Turing’s Law can sensibly contend that a computer can simulate any physically possible environment.
The improvement in the price/performance ratio of computer power is also unprecedented in human history. Moore’e Law is only the best-known measure of this. Since 1980, the cost per computation has fallen at a compound annual rate of 64 per cent.
That means the Cray-1 supercomputer, which cost the US government $8.8 million when it was the fastest computer on the planet way back in 1976, would now cost, to all intents and purposes, nothing.
Indeed, a computer many thousands of times more powerful than Cry-1 would also cost next to nothing.
Quantum computing, when it happens, will increase that power many times over.
So computers are undoubtedly a powerful technology already, and are becoming steadily more powerful as the future approaches.
How long does it take for a technology to transform an economy?
The question is: Why are they not transforming economies?
Of course, it took time for electricity to make an economic impact. It was not until 50 or 60 years after large scale electricity generation began in the 1880s that industrial production and household activities were transformed by electricity.
It took time and money to build generating capacity and national grids. It took engineers a while to design and build engines small and cheap enough to support household devices such as vacuum cleaners, radio sets and televisions.
It took manufacturing companies decades to replace steam-drive shafts and belts manned by unskilled workers with factories built as production lines, on which every operator had their own electric machine, illuminated by electric light, and had more expensive skills.
But once electricity was properly diffused through economies it transformed not just energy and manufacturing but food production and distribution (agricultural mechanisation and refrigeration), transport (electric trains), communications (telephony, the Internet) and household drudgery (heating, light, cooking, washing and cleaning).
If computing follows a similar 50-60 year time-lag to electricity, a great boost to productivity and living standards ought to be coming through soon.
After all, the first (mainframe) computers were deployed in the 1960s, the personal computer turns 40 this year, and even the first web server dates back 30 years.
Dramatic change is certainly needed if the digital age is to match the age of electricity.
Outside some narrow effects on telephony, media and entertainment, computing has not had a transformative impact in any part of the economy.
Unlike electricity, computing is so far a general purpose technology that has left swathes of economic activity untouched.
In fact, the first 60 years of commercial computing may actually have retarded productivity by increasing the volume of useless activity and distraction at work, multiplying the number and length of holidays and especially by drowning everybody in voices, words and images.
Is data fulfilling at last the destiny of computer technology?
But there are signs now that the trivial nature of the impact of computing to date may be no more than the necessary prelude to some genuinely revolutionary effects.
What was missing was the raw material for computers to process: data.
The prodigious memory and processing power of computers is dazzling, but their power to transform entire economies lies not in hardware or in software but in data.
In short, technology is not the point of information technology. Information is. It would be tedious to invest much time here in parsing the difference between information and data.
What matters is that computers need data to achieve transformative effects – and it has taken time for data of sufficient quantity, quality and variety to be digitised.
The Internet, email, mobile telephones, digital platforms such as Amazon and Uber, transport networks and now the Internet of Things (IoT) are producing vast and accelerating quantities of (mostly unstructured) data.
Business has so far forced data to fit into existing business models
So far, business has fitted this explosion of data into existing business models.
Companies have sought to capture and own data, or copyright it, and use it to sell more products and services or sell or license it to third parties.
Even Facebook and Google, in using data to sell the attention of consumers to third parties, are pursuing a surprisingly old-fashioned business model: it is comparable with television advertising in the days of audience surveys, just with more and better data.
So far, data has, in effect, reinforced the status quo.
Yet the fact that the business models of Facebook and Google would either collapse or have to change dramatically if they were obliged to pay users of their services for the valuable data they create provides a vivid insight into the power of data to challenge powerful commercial incumbents and overthrow existing ways of working.
After all, the amount of data created by consumers in a digitised economy is vast. It includes activities on Facebook, searches on Google, uses of apps, purchases on Amazon, viewings on Netflix, postings on YouTube, emails sent, usage of transport networks, financial transactions, health records, holidays, tax payments, passports and drivers’ licences, criminal records, jobs held, texts published, videos recorded, mobile telephone calls made and so on – and on.
At present this data is trapped inside private companies and government bureaucracies, largely in non-standardised and unstructured formats.
This enables companies to capture a part of its value, either by using it to sell more to existing customers or to attract new customers (as supermarkets and Amazon do) or by selling the data or the knowledge created by the data to third parties (as Facebook and Google do).
Consumer ownership of data could transform the current economic structure
If instead consumers owned all the data they create through their commercial transactions and their transactions with the State, companies would no longer be able to capture all the value of the data they hold.
Indeed, they would have to radically re-shape their business model.
Instead of a supermarket or Amazon or a price comparison website using the data it holds about a consumer to propose purchases – the sort of merchandising that a fishmonger in ancient Pompeii, let alone Gordon Selfridge, would understand – the consumer would initiate an auction to supply what they wanted or needed.
Companies would have to compete, continuously, for business, in the high street or the shopping mall as well as on-line, by alerting consumers in the market for a particular good or service even as they pass the store.
Instead of holding stock, manufacturers would have to manufacture to order.
Once an item was manufactured, a second auction would follow to determine the price to deliver whatever it is to wherever it is needed.
Eventually, as products are fabricated in the home by machines to data-led personal specifications, distribution networks will have to shift away from finished goods towards raw materials, which manufacturers will never be able to control.
To make money, “manufacturers” must have superior designs or engineering. To get paid, they will have to rely on automated invoicing and payment trails.
To initiate the auction, and complete the transaction, the consumer would share the minimum quotient of the data they own that is necessary to get the business done. This would include checking their identity and address, and that they have the means to pay.
By sharing even more data, consumers could ensure that goods or services are personalised, putting companies under pressure to de-standardise their offerings.
In services, switching between providers on the basis of price and consumption data can be coded and occur automatically and continuously. The customer “churn” that all companies detest today, is likely to become a way of business life.
The most popular business idea of today – turning every customer into a monthly subscriber – may well turn out to be the last gasp of an obsolescent corporate regime.
Open data is starting to make an impact in financial services
These effects are becoming visible in financial services already.
Open Banking is in theory (if not in practice) making it easier to change bank account, and the same model is being extended to insurance, asset management, retirement saving and mortgages.
Insurance is available now to cover the time that an individual is actually driving a car. Real-time discounts for good cars (Volvo is advertising motor cars with technology that activates the brakes when a pedestrian walks in front of the car) and good drivers are in prospect.
Likewise, the crude health insurance discounts of today (based on FitBits and gym visits) will be driven by data from heart monitors and grocery bills, and a reading of the genetic histories of families.
In banking, platforms are emerging that allow lenders to bid for assets, such as loans to small businesses and mortgages to householders.
The Open Banking standards initiative in the United Kingdom has attracted 301 firms to provide services based on data-sharing already, with another 450 queuing up.
What is needed to make the transformation happen?
An important question is how to make these developments happen faster and more widely, and in more markets.
It will almost certainly take a change in the law beyond the General Data Protection Regulation (GDPR) to give consumers full ownership rights over their data (beyond the consent, access and portability rights written into GDPR).
This should not be difficult, though it may take time. There is widespread recognition in government and the law that the corpus of commercial law needs updating to take account of new technological and commercial realities (and possibilities).
The harder challenge will be to overcome consumer apathy.
Most consumers are content to trade their data for “free” services, partly because paying for anything is irksome and partly because finding alternatives is tiresome.
Turning them into the activist consumers transformational change demands may be challenging.
It will be even more challenging if they are expected to appoint a trusted guardian – the banks, oddly, are judged best suited to play this role – to safekeep their data and manage their data requests.
Some consumers might prefer this option. A minority might even be prepared to pay a premium to a business capable of safeguarding their data.
But the vast majority of active, data-driven consumers must be created by an Artificial Intelligence (AI) that reads the data continuously to identify occasions to transact, wherever the data is held, makes it available to whoever needs to see it to complete the transaction, and checks the means are available to pay – while protecting confidentiality and privacy.
In this model, nothing more is required of consumers than accession to what the AI suggests. It should work.
As Open Banking says of its experience over the last three years, “if you provide people with the means to safely and securely put their data to use in their best interests, then they will do precisely that.”
How a transformed economy can boost productivity
A final question is: why would such an outcome boost productivity?
The single most important reason for believing a data-driven economy can lift productivity is that consumer ownership of data will break the oligopolistic (and occasionally monopolistic) structure of modern business.
Many incumbents will fail, taking their obsolete methods and capital with them.
Those that survive will be forced to invest in new equipment and processes, or risk being driven out of business themselves by new entrants armed with superior technologies.
Consumer ownership of data will spark waves of industrial restructuring and innovation akin to the introduction of the production line.
Competition driven by data will also put a premium on genuine innovation, as opposed to the current practice of incremental innovation designed to increase sales through product obsolescence.
Because data will provide an unending poll on ever-evolving consumer wants and needs and continuous insights into the best industries and opportunities in which to invest for the future.
More prosaically, the amount of manual processing will fall. Transaction costs, which are the largest single source of unnecessary cost in a modern economy – including distribution costs – will also come down.
Finally, data processed by AI has a natural propensity to yield scale economies: more customers produce more data that improves sales and so attracts more customers, creating a self-reinforcing cycle.
The challenge will be to prevent large corporations using that cycle to monopolise technologies, or form oligopolies, and raise prices, reduce output and suppress innovation.
The market shares assembled by the data incumbents (Facebook and Google) are discouraging in this respect.
The only solution is an effective competition policy that ensures data (if not the AI that processes it) remains open.
Written by Dominic Hobson February 2021