Is the Bank of England wrong about the robots?

It’s nearly a year since I delivered a TEDx talk on the question of whether robots will steal our jobs. Today’s news is that Andy Haldene, the chief economist of the Bank of England (BoE), has suggested that the UK will need a skills revolution to avert disaster as people are made “technologically unemployed” as result of AI shows how very little the debate has moved on since then.

The chair of the UK’s new Artificial Intelligence Council (AIC), Tabitha Goldstaub, has also emphasised that the goal is to create new jobs to replace those that are to be lost. While this might sound like a common-sense response to the issue – in this article I’ll explain why actually the BoE and AIC are together wrong about what to do about the robots.

Will Robots Steal our Jobs?

Yes. Emphatically yes.

Commentators make all sorts of complicated points to explain why the so-called ‘Fourth Industrial Revolution’ is really about to begin. Mostly their arguments revolve around the idea of ‘exponential growth’, which is an unnecessary fallacy that doesn’t actually explain the issue at hand.

As I explained in my TEDx talk last year, there is a simple explanation as to why robots will steal our jobs and which jobs they will come after first. The reason is simply that we are building machines today with human-level skills and abilities.

In the past, technology was designed to achieve things which people could not do without it. A hammer could exert more force than a fist. A spear could kill from a greater distance. A crane could lift more, a cart could carry more. A computer could count more, and more accurately too!

The new abilities that machines with super-human skills give us, was one of the primary ingredients for the Industrial Revolution. Another ingredient was the notion of breaking down a job into a series of tasks, and then combining man and machine (or woman and child and machine, as was often the case) in order to maximise an efficient production line.

Economists have spent the past few hundred years arguing about the effects of these two forces – technology and the division of labour. In fact, the opening sentence of Adam Smith’s great work The Wealth of Nations (1776) makes the case that it is the division of labour which is the greatest innovation:

“The greatest improvement in the productive powers of labour, and the greater part of the skill, dexterity, and judgement with which it is anywhere directed, or applied, seem to have been the effects of the division of labour”.

You might therefore be forgiven to have assumed that these two factors – technology being able to achieve ever more powerful things, and the division of labour, have led to a largely continuous process over the past few hundred years of industrialisation, and will continue to progress in the same linear fashion.

Such an assumption is easy to justify, as from the Western European perspective, writing this in late 2018, aside from some paradigm-shifting technologies (mobile telephones, email, the fax), the workplace of 50 years ago is broadly recognisable to that of today.

But such an assumption, that progress is slow and linear is a dangerous one to make – and plain wrong, for two reasons:

1)   Tomorrow’s technology will have more impact in replicating human skills and abilities than surpassing them

2)   Software is eating the world

Let’s look at each in turn.

Replicating Human Skills and Abilities

It’s been a fascination of technologists since the dawn of civilisation to build automata that looks like we do. Any techie who has ever visited Venice will no doubt have marvelled at the automata adorning the Torre dell’Orologio in St Mark’s Square whose various figures have been striking the hours and performing simple theatrical displays since the end of the 15th Century.

While we’ve been building technology to look like us convincingly for a very long time, replicating our own capabilities in machines has eluded us until now. Even at the end of the last Century it was thought that speech recognition might be a feat beyond our technological grasp for decades to come. The ability to analyse an image and write a descriptive summary of what was in the picture might be something that would take longer. The idea that we would build robots with dexterity such as the achievements of London-based Shadow Robot Company, or to make appointments on behalf of the owner such as Google’s recently unveiled Duplex, was until very recently the stuff of science fiction.

This has got nothing to do with exponential growth. Yes, all this does depend on a vast ocean of data and computational resources that was unthinkable a few decades ago – but in reality what we are experiencing today is simply the result of trillions of investment dollars that have created a tsunami of innovation aimed at replicating human abilities in machines.

And while some will surf that wave successfully, it threatens to drown the rest of us.

In 2014, Andrew McAfee and Erik Brynjolfsson published The Second Machine Age which gave credence to claims of the impact of automation in the workplace. Since then many studies have been conducted, and while the findings vary widely, the consensus is that around 40% of jobs will be affected by automation within the next 40 years.

I was drawn to this study when I was previously the Head of Technology for Deutsche Bank’s Innovation Lab in London. Automation was high on our agenda. Not just was the technology coming of age, but the Bank had a very public need to cut costs – which while we dressed it up euphemistically as ‘addressing our cost to income ratio’ – it was just economist-speak for headcount reduction. Reading the various methodologies employed by the various studies I devised my own, and considered the impact of automation technology on a Bank such as DB.

Broadly speaking, in order to assess the potential impact of automation, you first need to analyse the various jobs in existence (looking at either the economy as a whole, your community, or organisation) and consider the types of tasks that those jobs comprise of. A Bank might have a ‘cashier’ as a class of job title, but within that role are buried many activities – only some of which actually pertain to the title itself. Using this example, the primary task of a cashier might be thought to count the money handed between the Bank and the customer and vice versa. A secondary (and often forgotten) task though is to provide human touch customer service. No ATM is likely to be able to do that task any time soon.

It's actually this very thought that led me to realise what the impact of automation truly is. If you break down any given job title, you’ll be able to quickly see those tasks which could be better performed by machine (super human skills) such as counting money fast and accurately. Then there are those tasks which require human skill – such as understanding a customer complaint and directing the customer to the most appropriate source of relief. Such human-level tasks are exactly those which are currently under threat. Finally, and most importantly, there are those tasks which require human-touch. Customer service might be one such category. Another might be regulatory oversight, or creative roles in marketing. Many of these ‘human-touch’ activities are hard to precisely define, and deciding the line to be drawn will be, I believe, the skill of the leaders of the future – but what is shocking is that majority human-touch roles simply don’t exist. In an organisation such as DB who when I left employed around 100,000 people globally, maybe 80% of the roles comprise so little human-touch necessity that their automation isn’t just inevitable but imminent.

So 40% over 40 years is a useful soundbite, but it actually disguises the truth – the workplace of the future will be much more skewed than the one of today in terms of workforce. Some professions such as hairdressing will be unaffected by automation (it is a form of therapy as much as it is about the aesthetic), others such as lawyers will be decimated.

This is such an important point, I need to underline it:

1.    Every job can be broken into tasks

2.    Each of these tasks can be assessed in to one of three categories:

a.    Super-human skills are desired

b.    Human-level skills are desired

c.    Human-touch skills are desired

3.    Only those jobs with a high proportion of human-touch dense tasks will avoid automation

I could end the article on this bomb-shell, but there is another equally important point to make, and ultimately the one which has been glossed over by the BoE and AIC today.

Software is Eating the World

In 2011, Marc Andreessen published an essay in the Wall Street Journal entitled Why Software is Eating the World. The title has gone on to be something of a meme, yet while Andreessen was really making the same case that Alan Turing did 75 years previously, that universal computers (what we now call Turing-complete) are much better than single-purpose devices, the phrase actually sums up the much more powerful force that Adam Smith had envisaged three centuries earlier.

One of the central tenets of software engineering is to take a problem, and break it down into component parts. Each of these parts should themselves be broken down into modules, and these modules likewise. Such atomisation of code ultimately leads to greater code re-use, higher reliability, as well as a Software Development Life-Cycle (SDLC) benefit of greater collaboration between developers who can each work on different aspects of the problem and yet push their particular part of the solution live without worrying (too much) about the whole. Microservices and DevOps are the buzzwords of the day, but they are just new ways of doing exactly what Adam Smith envisaged.

This software engineering mindset is as powerful as it is dangerous.

As a way of tackling monolithic Enterprise applications of old, and turning them into highly scalable open-source SaaS solutions, it’s a winning formula. As a way of destroying the real-world it’s equally as effective.

The so-called ‘gig-economy’ is a by-product of the SDLC. Take any real-world problem, such as what to eat tonight, and break it into the component parts. One part is to decide what cuisine to eat (Indian/ Italian/Chinese etc), the next is to choose the dishes to order, then the purchase, and then the rather inconvenient logistics exercise of getting the food picked up or delivered.

Companies such as Just-Eat or Uber-Eats have architected this to perfection. By breaking down the process and recombining it into a single app they have in turn created a marketplace whereby previously independent delivery firms can focus on what they are good at (cooking) at the expense of what they weren’t (marketing, commercials, and logistics). The app vendor does the high-value marketing and payment exercise (which is where all the Capital Value is created), and outsources the highly inconvenient logistics process to anyone who is willing to pedal the streets on minimum-wage (or less, as most are actually paid by the job and not by the hour).

The result isn’t just a win for the consumer, who get a vastly improved culinary experience, but for the app-vendor also, who get to circumvent the last two centuries of labour laws. If Werner Sombart were still alive, he’d have recognised this to be a new stage of Capitalism, unimaginable to him writing Der moderne Kapitalismus at the turn of the last century without the subsequent invention of software.

The so-called gig-economy has created a new class of worker who have been described as the ‘precariat’. On minimum hours contracts, encouraged to contract through Limited Company shells, they are devoid of employment rights. While it works for some who get to run highly tax efficient schemes, for many it is a new source of anxiety that economists have been slow to recognise, and politicians slow to respond to.

Of course, it’s easy to see through this lens that the combination of software eating the world in this way and automation technology is insidious. Replace the Uber-Eats driver with a self-driving car, and the take-away with a robot food-factory, and you have the perfect system.

Is Resistance Futile?

So if Software is Eating the world, and Robots are going to Steal our Jobs – then what to do about it?

Well, Tabitha Goldstaub is absolutely right to say that one of the challenges is to ensure people are ready for change. However, focussing on skills and jobs, as Haldene and Goldstaub have in the news today is futile.

Hans Moravec argued that the job-landscape is like the countryside, with mountains and valleys, lakes, and coast-line. Every year the sea-waters rise, and the coast slowly shrinks. People standing on the land look upwards at their local high-point as the waters rise, and say “let’s climb the skills hills” and slowly they start climbing. Eventually the waters rise so high that there are too many people standing on low-lying land with no-where to run to.

To borrow from Moravec’ metaphor, what we need to do isn’t to help people scale the skills hills, but to free them from this anxiety all together and build boats so they can float on the sea of Capital that is being created by the Silicon Valley super-rich.

I readily accept that it’s terrifying to imagine a world without jobs (or at least one without jobs for all of us). What’s more terrifying to me is not imagining solutions of how to deal with the problem when it comes. To simply attempt to open the ‘magic jobs drawer’ as Calum Chace puts it in his book The Two Singularities (2018) is foolish and misguided, albeit at least a step that gives comfort to those most anxious about the coming revolution.

To show real leadership however, we need to not simply passively adapt to this new stage of Capitalism that the SDLC has given us but consider the opportunities for society in the stage ahead. Post-Capitalism is a collection of ideas that imagine and speculate beyond our current economic paradigm. There is no consensus in this movement as to the desirable outcomes or even the appropriate steps to take, but what grounds all who subscribe to these ideas is the common view that our current economic trajectory is unsustainable.

What the UK needs therefore is not an Artificial Intelligence Council that is comprised of technologists and technocrats but one that comprises a much larger cross-section of society to decide what is the realm of human-touch activities that we want to preserve as the exclusive reserve of human-labour.

Above all, we need a Bank of England which doesn’t limit itself to Capitalism either in its old or present form as the solution to our economic problems. We should all beware of a Central Bank which does this, as a Central Bank that is in denial of the factors that cause a crisis just as much as one that is in denial of the levers to pull to avert one. As we have seen before, post-fact diagnosis might be an interesting intellectual exercise, but at the cost of the very same misery for millions that Haldene today argues we need to avoid.

Just how far is the Bank willing to act though to lead us through the robot revolution? How far are all of us willing to free our thoughts to imagine the world we want to create?

These questions can be asked only by ethicists and not economists.

And only answered by you.