Will robots “Exterminate!” jobs?

In December 2016 Mark Carney gave the Roscoe Lecture in Liverpool and spoke about the challenge of automation. He made the point that, “…every technological revolution mercilessly destroys jobs and livelihoods – and therefore identities – well before the new ones emerge.” His talk reflects growing concern around the world about the impact of robotics, machine learning and artificial intelligence. Some take an optimistic position arguing we can and should race “with” rather than “against” the robots, others are more concerned about the social and economic consequences. The following books reach differing conclusions,  but pretty much all agree on the accelerating pace of change and the ubiquitous impact they are set to have on the labour market.

In 2014, Erik Brynjolfsson and Andrew McAfee published “The Second Machine Age” It followed an earlier and much shorter work entitled “Race Against the Machines”. I start with this, much quoted, work as it provides a good overview of the brynspeed and scale of change. They start out with an interesting analogy to illustrate the accelerating pace of change in information technology based on the operation of Moore’s Law. That “law” was first enunciated by Gordon Moore in 1965. Over the years the shorthand version has become that roughly every two years the complexity of integrated circuits double and their cost halve. The critical consequence of this is the power and costs of computing double and halve respectively also.

This law has proven reasonably accurate over the past 50 years which means of course the power of computers has doubled 25 times since it was first described. Brynjolfsson and McAffee then use an analogy to give some indication of what may happen in the future if the law continues. They use the, possibly apocryphal, story of the reward provided to the man who invented chess. When asked by the emperor what he wanted he said he would like rice, and the amount of rice to be the exponential sum of rice if you started with one grain on the first square of a chessboard and then two grains on the next square and then doubled again on each remaining square of the chess board.

His request was granted until the amount started to become apparent. After 32 squares it amounted to 4 billion grains of rice which is about one large fields worth. So the first half of the chess board delivered a significant but manageable reward. The problem was when they moved into the second half of the board. Geometric progression really starts to accelerate and by the time you get to the 64th square you are at 18 quintilian grains of rice, more than has ever been grown on the planet. In passing, the emperor had the inventor’s head chopped off. No one likes a smart Alec.

The point that Brynjolfsson and McAfee are making is that computer technology is now approaching the second half of the chessboard and the increases in power which have been impressive to date, will pale into insignificance as we go forward. Computer power will increase at such a rate that it is difficult to predict what the limits of its capacity going forward will be or its consequences.

When you read Brynjolfsson and McAfee you cannot but be infected by their optimism. They are clearly impressed by the speed and depth of challenge technological change is creating. Despite the optimistic tone of their writing they do see the issues which this rapid change generates. In an article for the January 2016 World Economic Forum they recognised that “Globalisation and technological change may increase the wealth and efficiency of nations and the world at large but they will not work to everybody’s advantage…” They go on to make clear “ordinary workers” will bear the “brunt of the changes”.

Further they see the negative outcomes are likely to be increasing inequality of wealth and incomes leading to inequality of opportunity meaning that not all talent will be accessed, undermining the “social contract”which underpins social order. They also state, perhaps prophetically given what happened in the remainder of 2016, “Political power, meanwhile, often follows economic power, in this case undermining democracy.”

They argue, because of all this, there is a need for greater social investment to allow the provision of good quality basic services including education. They also make the case for public sector investment to boost the economy in the short term however they also talk about simultaneously putting in place a fiscal consolidation plan. There is a bit of cake and eat it here.

Coming from a rather different position is Martin Ford whose book “The Rise of the Robots” was the FT’s surprise business book of the year 2015. Ford makes the point that Information Technology is a radical “general purpose technology” by which he means it is similar to fordelectricity or steam. These advances had implications in every area of life. Electricity effected production, transport, communications, culture and opened the way to the creation of completely new industries. IT is a similarly powerful and ubiquitous innovation, what is more the pace of its evolution is accelerating across the second half of the chessboard.

Ford focuses on the economic impact of IT on employment. He describes how the declining costs of technology and the rising costs of labour are undermining the competitive advantage of developing economies workers. Foxconn, for example, who make Apple devices in China announced in 2012 plans to introduce up to a million robots into its factories there.

Another interesting example Ford presents relates to the textile industry. This was decimated in the US in the 1990’s as jobs went to low wage economies like China, India and Mexico. But between 2009 and 2012 US textile exports increased by 37% on the back of automation technology that could now compete with the lowest wage economies on the planet. Of course this does not mean the jobs that were lost have been recovered. It means the jobs were first exported to the developing economies and have now been replaced altogether by technology. Mr Trump might find it more difficult than he thinks when he tries to bring the jobs back home.

Of course many of the jobs that have been lost to technology to date have been precisely the manual jobs which provided reasonably well paid work for the lower skilled working class. With the development of machine learning and artificial intelligence however a wider and wider range of jobs are becoming vulnerable. The success of Deep Blue in beating Garry Kasparov in 1996 was seen as a major milestone in the development of computer capacity.

Whilst this feat was impressive it was largely a result of what might be called brute computation. There are an immense number of potential chess games however they are all derived from a set of basic moves which can be easily programmed. Advantage is obtained by computing the potential future moves given any particular configuration of the pieces. As computers become able to compute ever larger numbers they can see further ahead in the game which gives them the advantage.  In crude terms their sheer number crunching ability eventually gives them an unbeatable advantage. The success at the chess board is in large part a result of quantitative advances in IT.

Some argue we are now moving in to a period where qualitative changes are being made. It may be that the scale of quantitative change leads at some point to a qualitative change but different approaches are being made to the development of algorithms such that machine learning becomes self sustaining. It is the difference between programming a computer with chess moves on the one hand and enabling the computer to learn how to play chess itself. When this is combined with natural language communication some very impressive results ensue.

One of the most impressive of these is mentioned in many of the books. It is Watson, another IBM supercomputer, which managed to beat two champions of the American TV show Jeopardy in 2011. Jeopardy is an altogether different game from chess. It does not have an easily programmed set of rules. It is based on cryptic, natural language questions which may involve humour, slang, high culture, popular culture, deliberate red herrings etc. There is no straightforward set of basic moves you start from or algorithmic rules to move forward with. The Jeopardy victory was great publicity but the remarkable capacity of the machine was not developed to win a TV show.

Immediately after the show Watson was launched as a diagnostic tool in the health industry. By 2013 it was helping diagnose health issues and design patient treatment plans at major medical facilities in the US. IBM see Watson as having a wide range of applications based on natural language information requests. It is now starting to be used by companies to review business strategies. And for the avoidance of doubt I do not mean making sure their budget numbers add up. I mean taking a view about what products should be launched when, and where R&D expenditure should be made. Of corse Watson is not the only game in town.

Eureka is a programme which was set to work out the mathematical equations which explain the  motion of two pendulums, one dangling from the other. A fiendishly difficult problem, according to those that know about these issues, and incomprehensible to those of us that did woodwork. According to Ford the programme, “…only took a few hours to come up with a number of physical laws describing the movement of the pendulum – including Newton’s Second Law – and it was able to do this without being given any prior information or programming about physics or the laws of motion.”

The implications of this level of machine learning for the professions has been analysed by R Suskind and D Suskind in their rather academic, but non the less interesting book “The Future of the Professions”. In broad terms they see robotics and automation as mainly impacting on manual and administrative jobs. However software is now set to have the same transformative impact on professional jobs.

susskindIn true academic form they begin by carefully defining what a profession is and the model of the relationship between professionals and society. The definition talks about a body of specialist knowledge; admission through the gaining of credentials; a regulatory framework and a set of common values.They then identify the problems that the current model is experiencing. Economic issues of affordability, technological challenge in the manufacture and distribution of knowledge and a growing suspicion that professional knowledge is used to blind people with pseudo-science at the behest of those with economic power etc.

One of the most interesting parts of their work is the review of what is already happening. They systematically review, Health; Education; Law; Journalism; Management Consulting; Tax and Audit; Architecture and …Divinity! What is already happening is shocking but Susskind and Susskind make the same chessboard point. Technological innovation is on an exponential path and it will have the kind of transformative impact on the professions that digitalisation has had on the music industry.

Ford’s work reinforces the same points that automated sports coverage and news articles are starting to appear regularly, if not always identified as such. Precedent searches for law practices, journal reviews for doctors, automated diagnostic programmes are all examples of where technology is starting to make inroads. Whilst currently the focus is mainly on the effective analysis and review of vast quantities of data. It is replacing the need for some of the more routine middle class functions in the legal and health professions. It is a moot point whether a computer will take Silk but each year the possibility of this happening is becoming greater. Even if this never happens, a huge number of middle class jobs are at risk over the coming years, put that together with the more mundane jobs that are going and the economy looks set to need fewer and fewer people.

It is usually at this stage in the reviews of IT and AI progress that people start mentioning Luddites and the need for us to learn from history. The point is made that 300 years ago circa 90% of the population worked in agriculture. Now it is something less than 1% in the UK. As productivity in agriculture increased other opportunities opened up for the workforce that was thus “freed up”. Another, oft quoted, example is the development of the internal combustion engine. When this replaced horses millions of jobs were lost in breading, maintaining, and  cleaning up after them. However, millions of new jobs were created in the production, distribution and maintenance of cars. Given all this we should not worry too much about the IT revolution automating existing jobs, this will “free up’ labour to move into new jobs that will surely emerge.

Ford is not so sure about this, and neither am I. As computing power becomes ever more powerful and also ever cheaper robots will become ever more flexible and may well be able to take on whatever new jobs are created. There is a qualitative difference between previous technological revolutions and the current IT based one. Previous revolutions, brought about by general purpose technologies such as steam, electricity and the internal combustion engine, have been about introducing a source of brute power which humans use to magnify their efforts. Initially IT did something similar magnifying our physical capacity so that, for example, highly automated Amazon warehouses could replace a huge amount of manual labour. However IT also magnifies our intellectual power, indeed in terms of the storage, retrieval and manipulation of data machines have far exceeded man for decades. Whether at some point machines cross the Rubican of consciousness is in some ways irrelevant. There are millions of sedentary jobs which currently middle class workers do which are at risk. The internet of things linked to a machine which has the capacity to have a natural language “conversation”, whether they understand that in some conscious sense or not, means enormous amounts of intellectual work will disappear as far as humans are concerned.

Ford certainly feels we are moving into a new economic paradigm where substantial numbers of people will be redundant in the very strong sense that there labour and intellect is not needed, not just by a company or an industry but by the economy as a whole. This concern is shared by another interesting writer on the subject, Ryan Avent in his book “The Wealth of Humans”.

Avent disagrees with the view that the digital revolution is radically different from what has gone before. He argues the economic process is very much in line with what happened in the industrial revolution. At that time society had to make a trade-off between “…new and improved goods, services and experiences at lower costs in exchange for social and economic disruption.” (my emphasis) Avent looks at the continuities of economic processes between the industrial and the digital revolution. Specifically he looks at the concept of scarcity and how that determines cost. He sees, like Ford and others, that labour is becoming more and more abundant. In this context labour “finds itself settling for a shrinking share of income – and is increasingly irrelevant in the taking of important economic decisions.”

Given that labours bargaining power in the economy is limited and it no longer has the power of trade unions to artificially boost its scarcity to increase its bargaining power, it has to turn to aventpolitics to protect its position. If the existing political elites do not seem to be responding then it is increasingly likely labour will turn to “…radical political movements that offer the possibility of political expression and economic power.”

Avent charts the way the economy has evolved over the past few decades, and how, what he terms, social capital has increased in value. He defines social capital as “contextually dependent know-how, which is valuable when shared by a critical mass of people.” He makes the point that 80% of the value of Standard and Poor’s 500 companies is now “dark matter” or “the culture, incentives and tacit knowledge that makes a modern company tick.” In other words an enormous proportion of value is created socially. At a national level social capital is within institutions like, the rule of law. However, the rule of law isn’t a thing, it is an emergent property arising out of the willingness of a citizenry to accept certain processes that substantiate the rule of law in practice. As Avent sees it, the benefits of the collectively created social capital of the digital economy are going increasingly to the owners of financial capital. Further he thinks that “… this mismatch is a source of significant economic trouble.”

I started this article with an analogy about the geometric progression of computer power. We are now entering the second half of the chessboard and thus the pace of change in relation to AI, machine learning and robotics is likely to be even more spectacular than in the past. There is another issue that is picked up in some of the books above and illustrated by a story, again possibly apocryphal, about a visit by a trade union official to a modern Ford Motor plant. The official is being taken around by Henry Ford Junior. Mr Ford is extolling the virtues of the robots which produce the cars without the need for food breaks, trips to the toilet or holidays, 24/7 and 365 days of the year. The never complain or argue for higher pay. At the end of this paean for the robot the trade union official turned to Mr Ford and asked. “And how many cars do they buy?”.

During the enclosure movement it was said that sheep ate men. Some think now that software is eating men. I hope new jobs do come and they are like the physically and socially sustaining jobs that emerged in the post-war trente glorieuses. Jobs that enabled people to have a reasonable standard of living and a sense of their own worth as part of society. I have my doubts however and think Avent is probably right when he comments on how the the global economy has evolved in recent years  and concludes”…the hardest part in finding utopia is not the figuring out of how to produce more. We’ve managed that. The hard part is the redistribution.”

 

The Second Machine Age, E Brynjolfsson & A McAfee. Norton Press 2014.                                   The Rise of the Robots, M Ford. Oneworld Publications 2015.                                                             The Future of the Professions, R Susskind and D Susskind. Oxford University Press 2015.      The Wealth of Humans, R Avent. Allen Lane 2016