At BOOSTMENA, the last annual conference hosted by BECO, an audience member asked Bhavish Aggarwal – Founder and CEO of OlaCabs, the Indian ride-hailing competitor to Uber – whether driverless cars would ultimately destroy much-touted jobs that these ride-hailing companies were creating.
Bhavish’s reply was quite matter-of-fact and pointed to India’s lagging transportation infrastructure and how the forecasted slow introduction of driverless cars (and primarily in large urban centers) would mean that jobs generated by these companies would likely endure for several decades to come. This [reply] is very significant for us in the MENA region: Careem, the region’s premier ride-hailing Company, has announced it expects to create 1 million jobs by 2018 with its latest round of funding.
But I think this question, as many writers point out, is much wider and involves technology’s potential destruction of jobs across industries with the resulting pressures for a growing global population.
If we look at current literature and studies on the different industries which are exposed to a systematic substitution of human jobs via automation, the most researched are: i) manufacturing; ii) retail; iii) financial services; iv) education; v) health care; and, vi) transportation.
Let’s look at some of the data and effects on each of these industries:
In the US, Manufacturing jobs have gone from a peak of over 19 million jobs in 1977 to ca 12 million in 2015, according to the Bureau of Labor Statistics – that’s nearly 40% loss as per Figure 1 below. However, manufacturing output has actually more than doubled over the past three decades. So what’s going on?
Figure 1: Employment in Manufacturing sector in the US
According to POTUS Donald Trump, “it’s China”! But according to more “enlightened” sources, increases in manufacturing output are due to investment in automation and software.
A recent Bloomberg article quoted the findings of a Boston Consulting Group study that stated that employing a human welder in a factory in the U.S. costs about $25 per hour including benefits; that drops to just $8 per hour for a robot, including installation, operating costs and maintenance. By 2030, “the operating cost per hour for a robot doing similar welding tasks could plunge to as little as $2 when improvements in performance are factored in,” BCG said.
The biggest buyer of industrial robots today is the automotive industry, with ca 40% of total global output of the machines – and sales of industrial robots have increased +110% in the last decade alone to over 250,000 per annum, as per International Federation of Robotics. This trend is not exclusive to the US.
China, which is estimated to employ 100 million people in the manufacturing sector and accounts for roughly a quarter of the global manufacturing output (up from only 3% in 1990!) is keen on introducing “millions of robots” in their manufacturing plants in a bid to retake the lead on manufacturing cost -today it lags far behind its competitors in terms of the ratio of robots to workers: South Korea has 478 robots per 10,000 workers; in Japan the figure is 315; in Germany, 292, and in the US it’s 164. In China, that number is only 36. Nonetheless, China has already lost 15% of its manufacturing workforce or about 16 million jobs between 1995 and 2002 and this pattern is likely to accelerate in the future as the Chinese state drives its initiatives to make Chinese manufacturing plants more competitive. So, President Trump is likely half right – it may be led by China in the future, but it will be a global drive to automation in manufacturing that will erase jobs in that sector.
We all have seen the introduction of touch screens and conveyer belts in restaurants (Yo! Sushi in Dubai) but the potential for automation in restaurants will only increase: Momentum Machines of San Francisco, has built a burger-making machine capable of producing 360 gourmet hamburgers per hour – the Company argues that eliminating labor costs and reducing kitchen space required would allow fast-food restaurants to spend more on high-quality ingredients.
McDonald’s alone employs nearly 2 million workers in >35,000 locations across the globe today.
In Japan, Kura sushi restaurant chain employs almost no humans in its 262 locations. Customers order using touch screens, robot chefs make sushi and conveyer belts deliver and whisk away sushi to/from the tables. Kura’s automated-based business has allowed it to offer sushi plates at ca $1 – far below its competitors.
Amazon has been taking on not only book retailers since it was founded by Jeff Bezos in 1994, but increasingly traditional retailers. Prof Scott Galway of NY Stern School of BusinessUniversity discussed in his presentation at the DLD2015 conference how traditional retailers such as Macy’s were following Amazon’s lead of replacing high-paying sales jobs in their departments stores on Main St with lower-paid warehouse jobs in rural centers that served their online stores.
Today, having lead the charge in this shift to online sales (and now extending to groceries through its Amazon Go! Service), Amazon today is more valuable than all the other traditional retailers combined (including WalMart! See Table 1 below). While Amazon was only worth $17.5Bn in 2006, today it is worth over $350Bn – in the same time frame, the combined value of the traditional retailers has dropped from $400Bn to just under $300Bn. And if you’re wondering what the employment to revenue generation looks like: Walmart generated $482Bn of sales in its latest financial year with 2.3M employees while Amazon generated $107Bn of sales in same period with a 10th of the work force! That’s 2.3x more efficient than Walmart. But that’s not the end of the story – Amazon purchased Kiva Systems, a warehouse robotics company, in 2012 that automates warehouse operations. As of 2013, Amazon had introduced 1,400 robots into its warehouses. And closer to home, an Indian company called Grey Orange Robots is selling warehousing robotics to warehouse operators in the GCC. Yes, automation is coming to a warehouse near you as well!
Table 1: Valuation of Traditional Retailers vs Amazon 2016 vs 2006
On September 25, 1995, NYSE member Michael Einersen, who designed and developed the system that allowed traders to receive and execute orders electronically, executed 1,000 shares of IBM through his hand-held computer (HHC) ending a 203-year process of paper transactions and ushering in an era of automated trading. Today, over 80% of all stock trading volume in the US is carried out electronically, mostly via “algorithmic trading”. The impact on jobs on the largest exchange in the world has been dramatic: in 1980, there were 5,500 floor traders working on the floor of the exchange. Today, that number has dwindled to about 500.
The impact of automation in the financial services industry has been widespread, not only in trading houses. At the start of the last decade, Wall Street firms employed nearly 150,000 works in New York City; by 2013, this number had dropped to less than 100,000. In the UK, there’s been a similar trend. In spite of increasing profits in the past years, the total number of job cuts for Lloyds, HSBC, RBS and Barclays amounted to over 189,000 between 2010 and 2015.
Machine-based test marking is not new and has been used at least since I can remember going to school in the early 1980s. However, machine-based essay marking draws on advanced AI techniques similar to the ones behind Google’s online language translation. In 2013, a nation-wide petition aptly named, “Professionals Against Machine Scoring of Student Essays” garnered the support of more than 4,000 teachers and educators. The group argued that algorithmic grading is “simplistic, inaccurate, arbitrary and discriminatory” and that “a device cannot read”. Martin Ford, in The Rise of the Robots, argues that the fact that machines are unable to read is beside the point, “techniques based on the analysis of statistical correlations very often match or even outperform the best human efforts”, as a study by the University of Akron’s College of Education which involved 16,000 pre-marked essays from public schools, compared machine markings with those awarded by human instructors and found that the technology achieved nearly identical levels of accuracy and at times, the software being more reliable. Although the education establishment has been able to successfully resist the introduction of broad automation in its institutions; the need to bring down the cost of education will surely accelerate the drive to increased automation.
The launch and growth of MOOCs (massive open online courses) will surely also increase the pervasiveness of Internet-based building blocks to reform the education sector, the most high-profile example being the artificial intelligence course offered by Prof. Sebastian Thrun and Peter Norvig at Stanford in summer of 2011. By the time the New York Times wrote a piece on the course in August of 2011, enrollment had grown to 160,000 from 190 countries. In “Lo and Behold”, Werner Herzog’s documentary on the rise of Information Technologies since the 1960’s, Sebastian Thrun states that within that class, the top 100 students had all registered remotely, which opens up a broader discourse about democratization of education and technology-aided reach but also about the degree of competition that we will see in the future.
Closer to home, a series of programs and start-ups have launched MOOCs in Arabic across our region: from the Queen Rania Foundation’s “Edraak” platform that offers courses from Harvard-MIT Consortium’s EdX, AUB, AUC and Bayt.com in fields as diverse as Communication Skills to Statistics & Epidemiology in Public Health; to Rwaq, a Saudi-based startup that offers two dozen courses online in Arabic only; and, in Lebanon, MenaVersity, a local start-up has also launched a variety of free online courses. Most of these platforms have more of an economic and social development remit, but as the business model switches to monetizing content, the cost of education (and ergo the staff numbers required to provide it) will come down.
5. Health Care & Life
IBM has had a long history of investing in projects it calls “grand challenges”, in which the Company showcases its technology expertise; one such “grand challenge” was the development of “Deep Blue”, a supercomputer that beat world-champion chess player Gary Kasparov in 1997. Another such challenge was to develop a program capable of beating the 50-time Jeopardy! champion Ken Jennings. The program that finally prevailed was called Watson, a massive collection of information and thousands of separate algorithms that it launches to search, compare, analyze and translate the best answers for a particular question with a given clue. Shortly after beating Ken Jennings and Brad Rutter in two televised matches in February 2011, IBM began to repurpose Watson for real world problems in medical diagnosis – it offers medical practitioners the ability extract precise answers from millions of references in textbooks, medical journals, clinical studies and even doctors’ and nurses’ notes. More and more leading medical institutions such as the Cleveland Clinic and the MD Anderson Cancer Center in Houston leverage Watson’s powerful algorithms to diagnose problems and refine patient treatment plans. Going beyond “big data”, a new feature in Watson called “WatsonPaths” allows researchers and medical students to see the resources consulted, the logic used in the evaluation and the inferences it made to generate the answers, thus expanding the diagnostic techniques.
Technology could be leveraged to solve a growing demographic problem in health care in the developed countries: as life expectancy increases, the percentage of the population over 65 that requires care on a more regular basis also increases and this coupled with retiring physicians, exacerbates the problem. In the US, a study by the Association of American Medical Colleges predicts that the country will face a shortage of up to 95,000 physicians in both primary and specialty care by 2025. In the UK, the NHS projects a shortage of ca. 10,000 doctors by 2020. Globally, the situation is even more dire: the World Health Organisation (WHO) advises that the industry will be short 12.9 million healthcare workers globally by 2035. Today, that figure stands at a shortage of 7.2 million workers globally, but in 57 countries, the situation is deemed to be in crisis, with fewer than 2.3 nurses, doctors and midwives for every 1,000 people – too few to deliver the basic level of care needed.
Martin Ford, in an article in the Washington Post in 2011, suggested there may be a way to leverage AI to short-circuit the traditional 5-10 year program to train medical professionals: a four-year undergraduate or master’s degree, primarily trained to interact with and examine patients and then convey that information into a standard diagnostic and treatment system. These new practitioners, trained to augment their knowledge by utilizing a standardized AI system could handle routine cases, while referring patients who require more specialized care to physicians. This would be one avenue for technology to actually help create employment and expand the availability and reach of health care across the globe.
However, technology will also affect those parts of the health care provision chain where human costs can be reduced through automation; in Dubai, for instance, the Dubai Health Authority recently launched the first robot pharmacy at Rashid Hospital. The goal of this initiative, as stated by the DHA is to “achieve a healthy and happy society” and in doing so reduce “human error” and “waiting times in the dispensing process” (the robot can dispense up to 12 medicines a minute and manage 35,000 SKUs).
And now, my personal favourite: transportation.
If you haven’t heard that human drivers will be replaced by autonomous vehicles within our lifetime then you probably have been living in a cave for the past 2 years as this is a topic which seems to be constantly in the news. The big argument for driverless vehicles in transportation is the reduction in human error and consequently, human deaths. The numbers are quite staggering: in 2015, there were nearly 40,000 deaths on US roads with 4.4 million traffic accident-related injuries, as estimated by the National Safety Council. In the UK, the death toll was ca. 1,800 in 2014 with close to 200,000 injuries. In the UAE, that number (road deaths) was 675 in 2015 with close to 7,000 traffic accident-related injuries. The common denominator, according to statistics is clear: drivers’ attention span or better, lack thereof.
So, in essence, driverless cars will make roads safer for humans, and therefore the implication is that less humans will be required in the transportation industry. In a recent video footage of an accident caught on a Tesla’s dashcam, the viewer can hear the sound alert from the lidar warning of imminent collision prior to the accident actually taking place. This is because today’s lidar technology is so advanced it goes around objects (vehicles) to see what’s in front of them. Nonetheless, Tesla and other automotive manufacturers still recommend that drivers keep their hands on the wheel while in “auto-pilot” mode, which takes us to the question of liability and how regulators will draft policy given the complexity of identifying liability if there’s machine/instrument error at fault in an accident. Given that this is still unchartered territory, the introduction of driverless cars will probably require clearing many regulatory hurdles and social stigmas before becoming mainstream. I wrote a blog post back in June 2016 that provided estimates on the timeline for the introduction of autonomous vehicles according to a panel of experts, but the question really is “when” not “if”.
In the popular YouTube video, “Humans Need Not Apply”, the authors forecast that the 3 million jobs in the US transportation industry and the 70 million world-wide are basically “over” in the next decade. The argument goes, human driver-related costs (payroll, taxes, licensing, insurance, accidents, etc.) are 2/3 of most transportation companies’ cost base. So, replacing human drivers makes economic sense. The “next decade” seems a little extreme, especially if you consider the average age of car fleets in cities like Cairo, where the average age is 25. It’ll probably take 25 years from the date of introduction of autonomous vehicles for a city like Cairo to complete turn their fleet and in the meantime, I’m sure there will be lots of kinks that will need to be ironed out, so to speak.
But if you follow the driverless cars argument further and assume that in fact, driverless cars will be constantly on demand and in perpetual motion, then two things can be derived from this line of thought: a) end users will likely cease to own personal vehicles and will just share available capacity; and, b) cities will be transformed as there will probably be no need for parking lots, garages and repair shops. And in space-constrained cities like Cairo, this has big implications: jobs which are linked to the automotive industry are definitely at risk.
What’s happening and is common across all these industries is that as digital technologies advance, the coordination costs associated with handling complex products (think trading of futures and M&A mandates in Financial Services; diagnosis in health care; specific itineraries/routes in transportation and logistics, etc.) decrease thus making market-based transactions feasible (instead of hierarchy-based) for a larger set of activities. Prof. Sundararajan explains this in his book, “The Sharing Economy” as follows:
“The way economic activity is organized is based on the relative magnitude of production costs and costs of coordinating different activities through the market. Let’s call the latter external coordination costs. When external coordination costs are low relative to production costs, this favors organizing economic activity through the market, where an individual will simply make products to sell directly to other individuals. If the opposite holds true, it makes more sense to organize the activity within a firm or “hierarchy” (the dominant model over the 20th century as these organisms were best suited to integrate mass production and mass distribution which evolved from the industrial revolution in the 19th century).”
I’m sure we have all heard the term “the unbundling of …” [fill in the blank which could be hotels, print industry, the automobile, etc.] where traditional industries such as the hospitality industry which had been dominated by large hotel groups are being “unbundled” by competitors that attack certain parts of the value/distribution chain, for instance, Booking.com for initial booking; AirBnB for lodging, etc.
Whether automation of fundamental tasks in manufacturing or services, or the advent of “crowd-based capitalism” through the so-called platforms businesses mentioned above (AirBnB, Careem, Etsy, etc.) whereby assets are more efficiently used and shared, it all still points to the fact that there will likely be less jobs or said differently: less full-time jobs and we may end up having a series of income streams from a set of diversified activities throughout the day (driver, host, etc.).
So is it all doom and gloom? No, I don’t think so –as my partner, Dany Farha, recently wrote in a piece on Arabian Business titled Tech is the New Oil, knowledge- or tech-based jobs will have greater staying power as more jobs will have an integration of human-computer tasks and activities: think data scientists; the new AI-assisted medical practitioners posited by Martin Ford above, etc.
In general, the future of jobs is being debated along two camps: a) human work augmented/assisted by machines; and b) a so-called “Universal Basic Income” which guarantees a minimum income for the citizens of each country.
The latter entails a vision for our societies where people are replaced to a great extent by machines which generate the overwhelming share of Gross Domestic Product and this is in turn distributed on a per capita basis. This (I would argue) dystopian future sounds eerily similar to the future described by Kurt Vonnegut in his famous book, Player Piano where an elite “engineer class” watch over the machines that produce everything and the rest either become soldiers, part-time construction workers, bar & restaurant owners or simply sit at home.
Indeed, we need to be aware of what attributes will allow our children to secure the jobs of the future. This I discuss in a 2nd part to this blog post fully aware that the reader by now may be somewhat exhausted.
- Martin Ford, “The Rise of the Robots”
- Arun Sundararajan, “The Sharing Economy”
 “The People’s Robots”, Will Knight, MIT Technology Review, May/June 2016
 Lidar = light and distance ranging; a low-pulse laser technology used in driverless cars to detect objects and motion