Artificial Intelligence and Meaningful Work

“One of the most robust findings in the economics of happiness is that unemployment is highly damaging for people’s wellbeing. We find that this is true around the world.”

“Not only are the unemployed generally unhappier than those in work, but we also find that people generally do not adapt over time to becoming unemployed unlike their responses to many other shocks.”

Excerpts from “Happiness at work”, CentrePiece magazine, Autumn 2017, London School of Economics and Political Science


“Participating in the satisfying work of innovating enriches lives by endowing them with purpose, dignity, and the sheer joy of making progress in challenging endeavors. Imaginative problem-solving is part of human nature. Participating in it is essential to the good life – and no elite minority should have a monopoly on that.”

“The technologies our species is developing might either hold the keys to unlocking human potential — or to locking it up more tightly than ever.”

Excerpts from “Meaningful Work Should Not Be a Privilege of the Elite”, Harvard Business Review, 03 April, 2017


“Viewed narrowly, there seem to be almost as many definitions of intelligence as there were experts asked to define it.” – “Abilities Are Forms of Developing Expertise”, Robert J. Stenberg, Educational Researcher, April, 1998


 One day the AIs are going to look back on us the same way we look at fossil skeletons on the plains of Africa. An upright ape living in dust with crude language and tools, all set for extinction.” – Ex Machina (2014)


In Matilda, the children’s fantasy movie directed by Danny DeVito and based on Roald Dahl’s book of the same name, the lead character, six-and-a-half year old Matilda, on her first day of school correctly calculates the result of 13 times 379 in her head – much to the amazement of her teacher and classmates. If such an event had taken place in our classrooms, we too would have been astounded and many, if not all, of us would have described young Matilda as being intelligent or even a genius. Yet if we told you that we have a machine that can solve the very same problem in less than a nanosecond, would any of you describe it as being intelligent? We suspect not; although, some may describe the person who designed the machine as being intelligent.

What then is intelligence?

We conducted an informal experiment (read: an impromptu poll on Whatsapp) involving our school friends. We asked our friends a simple question: who was the most intelligent person in our year group at school? The experiment covered three different schools from three different cities. Without exception, the choice was unanimous for each school.

After collecting their responses, we asked each of our friends a follow-up question: what is intelligence? Some gave us the Oxford Dictionary definition, others referenced IQ or some other standardised test scores while others still bifurcated intelligence into ‘book smarts’ and ‘street smarts’. Each person’s interpretation of what intelligence is was somewhat unique. Despite that, each person came to the same conclusion of who the most intelligent person at school was.

Exploring some of the academic research available on the understanding of intelligence, we found that opinion of what intelligence is was just as, if not more, divided amongst academics and researchers.

While we, collectively as a race, may not have a unified understanding of what intelligence is this has not impeded our shared progress. We, however, do not have the luxury of not understanding artificial intelligence, its possible evolution from here on out and what it means for the future of employment.

Millions of blue-collar manufacturing jobs have already been automated away by machines. This was the low hanging fruit for artificial intelligence. It has been widely accepted for well over a decade that technology would gradually replace workers in process-oriented roles where the objectives are well defined and the operating environment is controlled. The development of artificial intelligence, however, now threatens to automate away non-routine jobs across a vast number of industries. PricewaterhouseCoopers has predicted that 38 per cent of American jobs could be automated by 2030. They identify jobs in industries such as human health and social work, financial & insurance, education, mining & quarrying and public administration & defence as those with high level of susceptibility to automation. McKinsey Global Institute is even more apocalyptic as it estimates that as many as 800 million workers worldwide may lose their jobs to robots and automation by the year 2030.

A survey of history reveals that many new technologies have been sub-optimally utilised for years, sometimes even decades, before a more effective use of the technology has been discovered. If artificial intelligence is being sub-optimally utilised, it may not be the case for long. In May this year, Google unveiled its AutoML project, which is based on the concept of an artificial algorithm becoming the architect of another artificial intelligence algorithm without the need for a human engineer. Facebook has also started incorporating AutoML into parts of its architecture while Microsoft invited teams to compete in an AutoML implementation competition.

Why does AutoML matter?

Until now, engineers and developers have used trial and error to choose the best algorithm or set of algorithms to solve problems. After model selection engineers are also heavily engaged in the iterative process of optimising the algorithm and its parameters to the specific problem at hand. This entire process is resource intensive. It requires hundreds, if not thousands, of man hours and mind-boggling levels of computing power. As a consequence, costs of solving a single problem can run into the millions of dollars.

AutoML, on the other hand, automates the entire process of model selection and optimisation; saving computational capacity by not having to optimise and re-optimise models; and significantly reducing development time from weeks and months to days. AutoML capabilities will only grow over time and the complexity of problems it is able to solve is also likely to increase.

The progress of AutoML has been rapid. Take the case of AlphaGo, the Go playing artificial intelligence developed by Google’s DeepMind, which defeated the world’s number one human Go player. AlphaGo was a technological marvel with 48 artificial intelligence processors and data from thousands of Go matches built into it. It was no match for AutoML, however. DeepMind developed AlphaGo Zero an algorithm that was only given the rules of Go and then proceeded to teach itself and create an algorithm to play Go – all without any additional human input. AlphaGo Zero defeated Alpha Go at its own game only 40 days later. In fact, during a period of 72 hours, AlphaGo Zero beat the original by a margin of 100 to 0. What is even more startling is that the AlphaGo Zero only utilises 4 artificial intelligence processors – a 12 fold improvement over AlphaGo in terms of processing power requirement.

If the example of AlphaGo Zero is a peek into the future of non-routine, dynamic capabilities of artificial intelligence then the role of humans in the workplace is at risk of being marginalised to oversight and system refinement.

Adoption of artificial intelligence outside the technology sector remains limited. Few companies have deployed it at scale. However, the business case for artificial intelligence adoption is strong. A performance gap between early adopters of the technology and laggards will become increasingly evident over the coming years. A widening of this gap is likely to result in an adopt-or-die type of scenario for the laggards. And may even lead to a “winner-takes-all” type of environment where even second-best is not good enough to survive.

As adoption increases, the implications for the human workforce are likely to be far reaching. To quote Wired: “The AI threat isn’t Skynet. It’s the end of the middle class.” The threat of artificial intelligence is seen as being so grave that many have toyed with the idea of a universal basic income – a guaranteed living wage paid by government – as a possible solution should artificial intelligence result in widespread job losses for the middle class. But what of human dignity and the meaning we find at work in solving problems and in collaborating with our colleagues? And what of our right to pursue happiness if we can no longer fulfil our ambitions and aspirations but rather live from one government hand-out to the next?


Investment Perspective


“We are subject to the processes and trials of evolution, to the struggle for existence and the survival of the fittest to survive. If some of us seem to escape the strife or the trials it is because our group protects us; but that group itself must meet the tests of survival.

So the first biological lesson of history is that life is competition. Competition is not only the life of trade, it is the trade of life – peaceful when food abounds, violent when the mouths outrun the food.” – The Lessons of History (1968), by Will and Ariel Durant

Away from capital markets, the personal investment implications of the development in artificial intelligence are far reaching. While we can be accused of being pessimistic, we do not want to be ignorant to the challenges artificial intelligence poses. We understand and acknowledge not only the benefits the technology could deliver to businesses but also in solving problems humans have struggled with for decades and centuries. Artificial intelligence may one day help us overcome cancer or develop early warning systems for natural disasters – such possibilities excite us. The blind, unchecked development of artificial intelligence, on the other hand, scares us as it could one day tear through the social fabric that binds us together and spread the venom of protectionism across the globe.

For those of us with children, we have many difficult decisions to make and challenges to overcome in helping our children prepare for the world that awaits them. In our humble opinion, the risk-reward for teaching and learning foreign languages is skewed to the upside. While Google and others may develop tools to reduce the friction of translation, one can never truly understand a culture without understanding its language. In a world where nations are becoming increasingly inward looking, we need to increase our cross-cultural understanding and there will be, in our opinion, great reward for those that can facilitate such understanding. Learning a foreign language is the first and most critical step in this process.

We encourage all of you to learn and to encourage your children to learn at least one foreign language.

On the capital markets side, we reiterate our earlier call that we are at the beginning of a long-term secular trend towards automation and recommend positioning in a basket of automation and robotics related companies through the ROBO Global Robotics and Automation Index ETF ($ROBO).

Follow us on Twitter @lxvresearch


This post should not be considered as investment advice or a recommendation to purchase any particular security, strategy or investment product. References to specific securities and issuers are not intended to be, and should not be interpreted as, recommendations to purchase or sell such securities. Information contained herein has been obtained from sources believed to be reliable, but not guaranteed.  


The Case for a Pickup in US Inflation

“Given the right economic conditions, business will make substantial efforts to train workers. When the economy is moving along at a healthy pace and firms are eager to hire additional personnel, individuals with few qualifications begin to find opportunities.”

“Labor is the largest cost of the business sector. It has two determinants: employee compensation rates and worker productivity. When employee compensation rates increase, labor costs increase. When worker productivity increases, business pays less to get a job done. Both rising compensation rates and stagnating productivity in the United States have made critical contributions to inflation.”

– Excerpts from Profits and the Future of American Society, S Jay Levy and David A. Levy (1983)


“[I]nflation will remain rather limited as long as bad money, here the vellon, is still driving out the good silver money. For this means that the total money supply is scarcely changing.”

“[U]nexpected reversals of monetary policy seem to be the rule, especially when inflation accelerates, and if uninformed rulers try to react to consequences not foreseen by them. As a consequence, one can expect no damage from inflation in the real economy only as long as it remains small and smooth.”

– Excerpts from Monetary Regimes and Inflation, Peter Bernholz (2003)


Inflation principally comes in one of two forms:

  • Rising resource prices; or
  • Wage growth outpacing productivity growth.

Given the services-biased structure of the US economy, wage growth outpacing productivity growth has a far greater and more sustainable impact on US inflation than do rising resource prices. With wage growth in structural decline, inflation has remained tepid in the US despite the best efforts of policymakers.

Services as a Share of GDPServices ShareSource: The World Bank

During the recent Fed meeting, the Federal Open Market Committee downgraded its 2017-18 inflation forecasts lower. Despite this, Fed Chair Yellen argued that weak pricing pressures are transitory. We are in agreement with Janet Yellen and find that US inflation is on the cusp of turning sustainably higher.

To assess the prospects of US inflation on a look forward-basis we analyse the relationship between inflation, wage growth and productivity growth. As proxies, we use the annual change in US CPI for urban consumers to represent inflation, the US unit labour costs for the nonfarm business sector to represent wages and US output per hour for all persons for the nonfarm business sector as a measure of productivity.

Comparing inflation to wage growth less productivity growth, we find the relationship to have moderately positive correlation. The R-squared using quarterly data from Q4 1997 to Q2 2017 is 0.52. This relationship is much stronger during periods the two measures are trending, either positively or negatively.

Change in the Consumer Price Index vs. Wage Growth less Productivity GrowthCPI vs WG less PGSources: Bureau of Economic Analysis, Bureau of Labor Statistics

The differential between wage growth and productivity growth has been trendless since 2011, swinging from negative to positive and back on an almost quarterly basis. No wonder then that the inflation environment has remained benign, much to the frustration of the Fed, who has pulled out all the stops to fight deflationary tendencies within the economy.

Despite the seeming absence of inflation, we find that inflationary forces have been gathering steam since 2014. This has failed to show up in the headline data due to the outsized impact of a handful of industries caught up in downturns.

Based on data provided by the Bureau of Labor Statistics, wage growth has outpaced productivity growth across a majority of industries from 2014 through 2016. However, workers in the oil and gas extraction, media related and retail focused industries have suffered from declining wages and this has kept a lid on overall wage growth.

Annualised Wage Growth less Productivity Growth by Industry (2014 to 2016)WG less PG IndustrySource: Bureau of Labor Statistics

As the base effects of the negatively impacted industries unwind, we fully expect, headline inflation measures to turn up and begin to exceed consensus expectations. Giving further credence to our assertion is the tightness in the US labour market. The jobs opening rate and the number of small businesses identifying job opportunities as hard to fill are either at or near their highest levels since the turn of the century. At the same time, US U-3 unemployment is at its lowest level since 2000.

US Jobs Opening RateUS Job Openings RateSource: Bureau of Labor Statistics

 US Small Business Job Openings Hard to FillJobs hard to fillSource: National Federation of Independent Business

US U-3 Unemployment RateUS UnemploymentSource: Bureau of Labor Statistics

Historically, periods of labour market tightness when businesses are facing difficulty in filling job openings have preceded increasing wage growth. Comparing the US Small Business Job Openings Hard to Fill index to US wage growth lagged by one year, we find this to be the case up until the end of 2012. Since 2013, however, the relationship appears to no longer hold true. The number of businesses reporting job opportunities difficult to fill has been increasing while wage growth has remained largely absent.

Small Business Job Openings Hard to Fill vs. Wage Growth (Lagged One Year)Job Openings vs WGSources: Bureau of Labor Statistics, National Federation of Independent Business

Once again, the relationship is seemingly impaired at the headline level due to the outsized impact of a handful of industries. Based on data provided by the Bureau of Labor Statistics, wage growth has been positive across a majority of industries from 2014 through 2016. The oil and gas extraction industry, unsurprising given the collapse in the price of oil in 2014, has been a major drag on overall wage growth.

Annualised Wage Growth by Industry (2014 to 2016)Wage Growth IndustrySource: Bureau of Labor Statistics

Going forward, the oil and gas extraction industry should no longer be a drag on headline wage growth and may even have a positive impact on it if oil prices continue to increase. We therefore expect wage growth to pick up as businesses increasingly pay up or hire lower skilled labour and train them up to fill outstanding job openings.

Small Business Job Openings Hard to Fill vs. Capital Expenditure PlansJobs hard to fill vs Capex PlansSource: National Federation of Independent Business

The effects of the structural deflationary forces of globalisation, migration / labour mobility and declining trade union membership are also abating. The lion’s share of gains from outsourcing has already been realised. Politicians are increasingly pandering to populous movements and turning to protectionist policies, making labour migration far less frictionless. Trade unions have held very little appeal to younger workers that entered the workforce in recent years.

The inevitable corollary is the rising labour share of corporate profits will place increasing pressure on businesses to improve productivity. We therefore expect capital expenditures to pick up. Businesses will increasingly invest in automation and robotics to overcome the challenges of wage inflation and labour market tightness.

Small Business Capital Expenditure Plans vs. Productivity Growth (Lagged One Year)Capex plavs vs PGSources: Bureau of Labor Statistics, National Federation of Independent Business

Increased corporate spending will not only lead to improvements in productivity but to an upturn in the overall US business cycle. Capital investment has been the one missing ingredient in the US economic recovery since the Global Financial Crisis.  As businesses spend more, corporate profitability will pick up, which will lead to increased hiring and higher wages, which will feed into further investments into automation and robotics.

Our base case is, therefore, that the current US economic expansion will be the longest ever recorded. And the business cycle will only come to a turn after unemployment levels fall below 4%, inflation exceeds prevailing expectations and policymakers begin to respond to the unexpected consequences.


Investment Perspective

US median household income has been rising and we expect it to continue to rising as wage growth accelerates.

US Median Real Household IncomeUS Median Household IncomeSource: US Census Bureau

At the same time, US household balance sheets have been repaired with the household debt to disposable income ratio in decline since the Global Financial Crisis. More so, the debt service to disposable income ratio is at comfortable levels for US households. There is ample room for households to take on more debt, especially for the poorest households who are the likeliest to benefit as wage growth picks up.

US Household Debt to Disposable IncomeUS Household debt to disposable incomeSource: Bloomberg

 US Household Debt Service RatioUS Household DSRSource: Bloomberg

As poorer households’ disposable income increases, this cohort is more likely to increase consumption as opposed to increasing savings, especially when compared to upper-middle and upper class households. Poorer households shop at Walmart not Whole Foods. They eat at McDonald’s not Shake Shack. We expect retailers and quick service restaurants catering to lower and lower-middle income households to be amongst the greatest beneficiaries of higher wages. We are particularly bullish on the prospects of Walmart ($WMT).

Consider the relationship between US wage growth and $WMT revenue growth lagged by one year. The revenue growth measure does not adjust for store openings, corporate actions and other extraordinary events that may have occurred during intervening periods. Despite the lack of adjustments, this dirty measure has shown a strong relationship with wage growth.

$WMT’s revenue growth has flat lined in recent years as wage growth has been trendless. As wage growth picks up, we expect investors to increasingly come to recognise $WMT’s growth potential and rotate out of Amazon and into $WMT.

US Wage Growth vs. Walmart Revenue Growth (Lagged One year)WMT vs WGSources: Bureau of Labor Statistics, Bloomberg

A derivative of accelerating wage growth and labour market tightness is business’ increasing investment in automation and robotics. We are at the beginning of a long-term secular trend towards automation. Rather than picking winners at this early stage in the trend, we recommend positioning in a basket of automation and robotics related companies. The most obvious way to play this theme is the ROBO Global Robotics and Automation Index ETF ($ROBO).

Lastly, another derivative of accelerating wage growth is that the Fed is likely to increase interest rates at a faster pace in 2018 than currently anticipated by the market. We expect the short end of the curve to rise faster than the long-end, resulting in a classic bear flattening.

US Wage Growth vs. Effective Federal Funds RateEffective FFR vs WGSources: Bureau of Labor Statistics, Bloomberg

We are long $WMT, $ROBO and looking to get short the short-end of the Treasury yield curve.