The Imminent Threat of Technological Unemployment

Part 1:  Technological Unemployment

The world is on the brink of a technological revolution in the field of artificial intelligence (AI), the fruits of which will change society forever and usher in the fourth industrial revolution.  Automation technologies like self driving cars, smart robots, and disease diagnosing algorithms will improve and save millions of lives.  The downside is these advances will eliminate millions of jobs over a staggeringly short time period.

My goal in writing this is to convince you that we are on a path towards economic and social disaster given these rapidly approaching advances in automation.  I fear that the prescriptions that populist movements are proposing to reign in globalization, curtail immigration, and raise the minimum wage will only increase the onset of automation by increasing labor costs.  The recent populist backlash will pale in comparison to the political response when the robotic revolution takes hold in the next 5 years.

We are already seeing these effects today.

We are now inventing labor saving machines at a rate that is eliminating jobs faster than new ones can be created.  It is estimated that robotics will create 2 million high paying jobs by 2020, but in the process will eliminate over 7 million existing jobs across the 15 major developed and emerging economies.

For example, Foxconn, the company that makes computer chips for the iPhone recently replaced 60,000 workers with robots.  They have plans to replace their entire workforce with robots, and are targeting 30% automation by 2020.  Additionally, Adidas announced that it will move its shoe production from Asia to Germany by building a fully robot-operated shoe factory.

In a recent Fox Business interview, the former CEO of McDonald's, Ed Rensi, said "It's cheaper to buy a $35,000 robotic arm than it is to hire an employee who is inefficient, making $15 an hour bagging French fries."

Contrary to current political dogma, most US manufacturing jobs have been lost to technology, not trade agreements and globalization.  Over the last half century, manufacturing as a share of US GDP has remained constant, despite steady decreases in employment from nearly 25% to under 10% today.

This is exemplified by the recent decision by Carrier technologies to keep approximately 1000 jobs in their Indiana plant instead of transferring the jobs to Mexico to save on labor costs.  This is heralded as a victory for the working class, but the fact that Carrier also announced its plans to invest $16 million in automation technology begs the question:  how long will these jobs even exist?

The outlook is grim.

A 2013 Oxford study predicts that by 2050, 47% of US jobs that exist today will be computerized.  A UN report predicts that over two thirds of jobs in the developing world will be eliminated.  Larry Summers, former US Treasury secretary, predicts that by mid-century, over one third of US men aged 25-54 will be jobless.  For perspective, the unemployment rate in the US during the great depression at its peak was less than 25%.

This will have disastrous effects on the social fabric of our society.  Millions of disaffected workers, mostly men, will be unemployable and will likely never marry.  They will be primed for demagoguery, tribalism, and extreme forms of ethno-nationalism.  This has never worked out well in the past.

Books like "Rise of the Robots:  Technology and the Threat of a Jobless Future" by Martin Ford, and "Men Without Work" by Nicholas Eberstadt explore these effects in great detail.

This is not the first time that technology has reshaped the economy, but it's never happened this fast.

The Luddites tried to save their jobs by destroying cotton and woolen mill machinery at the onset of the industrial revolution, and similar groups have echoed the same sentiment.  In the 1800's, the agricultural sector employed over 90% of the US population, but out of the industrial revolution emerged technologies that reduced the need for human labor, and now agriculture employs less than 1% of the population.

These groups incorrectly viewed capitalism as a zero sum game, and failed to realize the immense and wide spread benefits that progress and even disruptive innovations would bring.

This time it's different.

Part 2:  Machine Intelligence

We are no longer creating dumb machines that have to be explicitly told what to do.  The current generation of machine intelligence can actually think, solve problems, and react to new scenarios on the fly.  They use a technique called deep learning that uses hierarchical layers of simulated neural networks. These networks are given a goal and then trained on large data sets until they can create intelligent outputs.  This approach is biologically inspired and is similar to how a child learns.

Consider that all job tasks can be broken down into four types of work:

Routine physical work
Routine cognitive work
Non-routine physical work
Non-routine cognitive work

Traditionally, dumb machines could only automate routine tasks.  A spreadsheet can perform millions of calculations a second, but it could not determine if a picture contained a dog or a cat.  A robot can make thousands of precise welds, but it can't pick up a bolt in a box of screws.  As a result, automation was historically limited to routine tasks, and we have seen declining wages for these types of jobs since the 1970's.

This is all changing.  Deep learning powered machine intelligence can now automate non-routine work.  

In the 2010's, we are currently in the process of automating non-routine cognitive tasks with improved AI assistants, facial recognition software, voice recognition, advanced wall street trading algorithms, and systems like IBM's Watson.  Even highly skilled white collar jobs like medical pathologist will have to deal with the threat of automation.  Here are several examples of what has already been accomplished.

A team of 5 AI researchers developed an algorithm that is better at predicting pathology cancer outcomes than a panel of top experts that had nearly 100 years experience between them.  The computer programmers had a combined zero years background in biology or medicine.

IBM's Watson recently saved a patient's life from Leukemia by making a diagnosis that a team of doctors could not do.  Watson compared the patients genetic abnormalities to 20 million research papers in 10 minutes.  Given the fact that over 8000 new medical research papers are published every day, it is impossible for researchers to read them all.

The Google translate mobile app allows you to translate voice, text, and images in real time.  You can actually hold your camera up to any printed text, and have it translate it in real time on your phone's screen.  This star trek like ability exists today, and it's a free download from the Play store.

Looking into the 2020's, we will likely automate a significant amount of non-routine physical tasks in the form of self driving cars and trucks, manufacturing robots, service robots, and delivery drones.

If the future economy can automate the majority of work, how will people get an income?  Let's first look at some specific examples of automation that exists today or will in the near future.

Part 3:  Self Driving Cars

I believe the arrival of self driving cars will drive this conversation to the forefront of our politics.  It will be the in-your-face example of technological unemployment that is both relatable and impactful enough to force a national and global conversation.  Ten thousand office workers that are replaced by an algorithm operating on a server bank isn't a story that you can get behind.  Two million unemployed angry truck drivers, most of whom own guns, on the other hand, is.

The transportation industry directly employs nearly 3 million truck and taxi drivers, and millions more indirectly through insurance companies, mechanics, diners, restaurants, highway patrols, and hospitals.  There are entire towns whose existence is dependent on highway intersections, and all of these jobs are all going away in the near future.

For comparison, the coal industry at its height, employed nearly 300,000 people in the 1920's.  This number has steadily declined by tens of thousands of jobs per decade to around 150,000 today.  If this is enough to warrant national political attention, what will the response be to the elimination of one million transportation jobs in a single decade?

This is not hyperbole.  Stage 5 automation largely already exists today.

Tesla recently shared two videos that show a Tesla Model S driving to work hands-free from driveway to parking lot, available here and here.  In the first video, the car navigates down the driveway, through a neighborhood, onto highways, it merges with traffic, onto off ramps, it stops for pedestrians, it drives cautiously around construction cones, and obeys all traffic signals and signs.  It even drops the driver off at the front door, and then parks itself.

This technology is only going to get better.  Its all powered by deep learning algorithms that are trained on input and sensor package data.  Every time a Tesla is driven by a human or in self driving mode, it uses that data to improve its driving ability for all other and future Teslas.  Combine this with the next generation of dedicated deep learning hardware coming from Nvidia, and this technology is going to rapidly improve.

In 2014, the CEO of Uber stated their objective is to have a fleet of self driving cars.  They are already testing self driving Ubers in Pittsburgh, partnering with Volvo and Carnegie Melon.  Uber has also expanded into self driving trucks.  It recently purchased the self driving truck startup, Otto, which just completed its first delivery of Budweiser beer from the brewery to a distribution facility.  In April, a similar feat was accomplished by a platoon of networked, self driving trucks that drove across the Netherlands.

Most of the major car companies are competing and investing billions into self driving technology.  Ford predicts it will have self driving cars in 5 years.  BMW is also targeting 2021.  Tesla is probably a little closer.  Either way, the technology to replace millions of jobs is going to be commercially available sooner than most people realize.

Part 4:  Walking Robots, Automated Factories, and Robot Chefs

In the same way that electricity revolutionized every industry, AI is set to do the same.  Algorithms, service kiosks, and the internet of things will replace vast number of jobs in factories, warehouses, field technicians, call centers, restaurants, retail centers, and law firms.

If you follow these advances, it seems like 2016 was a tipping point where every couple of months science fiction casually became reality.  Here are some examples from 2016.

In February, Boston Dynamics, released a video showing off their latest version of Atlas, their bipedal robot.  The video shows Atlas opening doors, picking up boxes, getting knocked down, standing back up, and it even walks on snow covered hills in the woods.  How easy would it be to convert Atlas into a warehouse robot that stocks shelves or assist on construction sites?

In April, a Google-owned robotics company called Schaft, released their own video showing a smaller, less human-like, but still bi-pedal robot that can walk up stairs, can carry heavy weights, and even walk along a pebble covered beach.  One can imagine a robot like this being used to meet the massive demand for home healthcare aids for rapidly aging populations in the US, Germany, and Japan.

Also in April, Kinema Systems, released a video of their robotic arm called Kinema Pick that utilizes machine vision to break down boxes in a pallet, and move them to a conveyor belt with minimal instruction.  It uses machine vision to automatically recognize a box, and does not require bar codes or RF labels.  Imagine how seamlessly this arm could be used to unload pallets from self driving trucks.

Amazon purchased Kiva robots in 2014, and currently use their pallet-bots in almost all of their warehouses.  These pallet robots move modular shelving units, and human pickers pull the merchandise from the shelves and put them into boxes.  The complexity is amazing, but its striking how few human workers are present in such a large warehouse. Amazon wants to further automate their warehouses using robotic pickers to further replace human workers, and even automate the delivery process using self driving trucks, drones, and even flying warehouses.

Amazon Go is aiming for a new type of grocery store that completely eliminates the need for checkout lines.  By combining machine vision and deep learning, the store will know whenever you place an item into your bag and automatically bill your account once you leave the store.  This will eliminate checkout lines entirely.  This technology will be disruptive to the retail industry, a sector that employs over 7 million cashier and sales people, and it will be launched in 2017.

Moley is developing a pair of robotics arms that will prepare meals for you in your own kitchen.  These arms will chop vegetables, hold the pot while it whisks the sauce, and even clean up afterwards.  This sounds like science fiction, but Moley is aiming for a 2017 product release.  Imagine the impact Moley will have on the restaurant industry.

Robots are even going to school to learn how to pick and sort small objects.  Google has a robot school where arrays of robotic arms learn hand-eye coordination to pick up objects through trial and error.  Once they master this, all future robots with similar hardware will share the same skills through a simple software update.

These advances will enable entire supply chains, factories, distribution networks, and retail centers to be nearly fully automated.  This is both amazing and terrifying.

Machine intelligence does not stop at robotics.  It can also be used to automate tasks that require thinking, vision, and even abstract thought.

The crown example of this is when AlphaGo, a computer made by Google's Deep Mind project, beat the 18-time world champion Lee Sedol earlier this year at the board game Go.  The striking fact in this accomplishment, is that there are more possible moves in the game of Go than there are atoms in the universe, making any brute force calculation impossible.  Just months ago, this feat was thought to be decades away if not impossible for an AI.

In order to beat the top human at Go, AlphaGo had to anticipate moves and think abstractly about new scenarios.  It achieved this through a combination of reinforcement learning and deep learning where it practiced against itself million times to refine its own neural network until it developed intuition to beat a master Go player.

The CEO of Google's AI project Deep Mind, Demis Hassabis, has stated their mission is two-fold:

Step 1.  Solve intelligence.

Step 2.  Use it to solve everything else.

It seems that they are well on their way towards step 1.  If he is successful, that will be the last thing humans ever need to invent, but that is a conversation for another day.

Part 5:  The human cost of technological unemployment

Imagine the following thought experiment:  In a corporatist dystopian future, corporations get to ignore all labor laws with a singular goal in mind:  to maximize profits.  In this future, employees live and work in their places of employment under horrid conditions.  They earn whatever the market determines their labor to be worth, with no regard to quality of life.  They are connected to IVs with the perfect blend of nutrients and stimulants to maximize productivity.

This vision is entirely unrealistic.  Not because of ethical or political barriers, but because using human workers will no longer be the best way to maximize productivity.  Robotic workers are fundamentally better than human workers in nearly every way.

Consider the following traits about robotic workers:  they can work 24 hours a day 7 days a week, will never ask for a paycheck, never call in sick, never go on vacation, can work in the dark, don't require air conditioning, will never form unions, never demand better working conditions, never go on strike, never quit to take a better job, never go on maternity leave, or never sue for discrimination.  It can take months or years to teach a human worker a new skill, but a software update that is downloaded in minutes could transfer skills learned from other robots across the world.

The pace that AI is being developed is faster than humans can learn new skills.  Even if we completely reform the high school education system to produce a wave of students with world class critical thinking skills, change the university structure to 3 year programs with an emphasis on work related skills, and supplement that with free 6 month retooling programs at thousands of community colleges, these efforts will still fall short in combating technological unemployment.

There is a day rapidly approaching where the labor market simply won't create enough jobs to employ a critical mass of people who want to work.  Tens of millions of people, through no fault of their own, will suffer in poverty.  This will have devastating effects on society.

US and European politics are already seeing a rise in demagoguery and toxic forms of conservative populism from the effects of globalization and immigration that has disproportionately hurt low skilled workers.  What will western democracy look like when mass technological unemployment is added into the equation?  The Wall Street Journal, The Guardian, and The Atlantic have all written about this.  The outlook is bleak.

These effects will not be limited to developed countries.  Factory automation could also eliminate the development pathway of low skilled manufacturing in developing countries, potentially trapping them economically and culturally as third world nations.

So what do we do about it?

Part 6:  Universal Basic Income

Smith, Friedman, nor Keynes never accounted for mechanical minds that can think and outperform the majority of humans in their economic models.  The sacred dogmas of their economic systems are becoming less relevant.  What demand information is there to communicate by price if transportation and manufacturing costs are close to the cost of raw materials and energy?  Without jobs, what mechanism is there to create and transfer wealth among citizens?  If a robot workforce is developed en mass, what manufacturing or service niche can a country fill?

I am not going to pretend that I am an economist or that I can predict what future societies will look like.  However, I think its clear that our current economic systems and the status quo has no way of dealing with the technological unemployment that the next generation of machine intelligence will bring.

If we stay on our current path, we could end up in a techno-feudalism dystopia with a handful of tech trillionaires that own all of the robots.  The policy prescriptions to reign in globalization will not work to combat technological unemployment.  For the masses, this will lead to economic insecurity and social unrest.

The only option I see is to share the wealth created by automation in the form of a universal basic income (UBI).

There are multiple forms and variations of UBI, but in general its defined as a system where all citizens of a country regularly receive an unconditional sum of money from the government, in addition to any income received from elsewhere.  There would be no strings attached to this money, and every individual could live their life in whatever way they want to.

Even before the threat of technological unemployment there have been both left-wing and right-wing arguments for UBI.  From the left, it would increase equality because it would minimize the effects of gender and racial discrimination.  While for the right, it would reduce bureaucracy and shrink the government because anybody with a social security card and a bank account would get the same monthly electronic deposit.  The conservative economist, Milton Friedman, argued for a similar system in the 1960's, although he called it a negative income tax.

The UBI future could usher in an artistic renaissance while also rejuvenating the entrepreneurial spirit the likes of which capitalism has never seen.  How many artists didn't create art because they needed a day job to pay rent?  How many entrepreneurs chose not to take the risk of starting a new business because they had a family to provide for?  When income is decoupled from employment, people will be able to pursue their own paths towards happiness.  When given economic security, creative people will create.

However, the transition to a UBI society will bring many challenges and it will be far from utopian.  Work provides structure and meaning for millions of people.  It could create a generation of couch potatoes who never learn the value of hard work and rarely leave the comforts of their virtual worlds.  UBI would be disruptive to our culture, the social contract, the existing welfare systems, and it could close borders as rich countries try to curtail immigration to minimize costs.  Our basic financial needs may be taken care of, but humans are status seeking social animals, and we will find new ways to seek status and hate each other.

One of the great challenges of the UBI-generation will be how they adjust culturally and emotionally to a post-work world.

One of the questions is how we would pay for such a system.  Charles Murray, from the conservative think tank, AEI, has discussed the implementation of UBI in both an article in the Wall Street Journal titled A Guaranteed Income for Every American and in an interview with Bill Kristol.  In Murray's vision, UBI would replace social security, medicare, medicaid, food stamps, supplemental security income, housing subsidies, every other kind of welfare and social-services program, as well as agricultural subsidies and corporate welfare.

In my opinion, enforcing existing tax laws is a good starting point.  The 2012 report estimated that $32 trillion is being held in off shore tax havens, and the recent revelations from the Panama Papers confirm that corporations are hiding trillions.

UBI is slowly gaining traction and entering the main stream conversation.  It's been discussed by the US Congress.  Elon Musk recently said there's a pretty good chance we end up with UBI.  In a recent interview, even President Obama acknowledged UBI will enter our political debate in the next 10-20 years.  There are several UBI pilot programs that will soon be underway in the Netherlands, Finland, England, Ontario, Canada and Oakland, USA.  Switzerland even held a vote to implement UBI, which lost, but still received 23% supporting votes.

I really hope that I am wrong in my outlook.  For hundreds of years capitalism has created massive amounts of wealth, prosperity, and improved the standards of livings for billions of individuals.  I hope that the classic dogma that innovation will always create a net gain of jobs holds true.  However, given the trends that I have outlined above, I don't see how this could possibly be true.

Unless you can think of something better, I think UBI is the only viable path forward.  2016 is too early to implement a UBI, but its a conversation that we have to have now because somewhere between 2020 and 2030, I believe it will become a necessity.  If we do pull the transition off, I hope our future looks more like the world envisioned in the final chapters of the novel Manna than it does in the Disney movie Wall-E.