The AI Unemployment Myth

I keep seeing articles supporting a Universal Basic Income on the grounds that robotics, and artificial intelligence, are going to take over most of the jobs currently done by people.  Aside from all the ways that a Universal Basic Income can destroy an economy, it is important for people to know that robotics and artificial intelligence are not going to cause massive levels of unemployment.  Automation does not make work ‘obsolete,’ and it never will.

The first time automation caused a mass panic was likely in ancient Sumeria, with the invention of the shovel.  Prior to the shovel, dirt had to be moved with nothing more than hands.  One can imagine what hand-shovelers must have thought when someone put a flat rock on a stick and could suddenly shovel as much as twenty or thirty people previously could, combined.  What would Sumeria do with all of its laborers when one person could move so much dir their own?

The answer, of course, is that the Sumerians could move far more dirt in total than they could before, and/or could move laborers to other tasks, such as making bricks, or plowing fields.  The shovel did not end the need to work, but enabled people to accomplish more work than they could before.  The shovel vastly improved their lives.

The steam shovel was another invention people thought would make work obsolete.  The car was another.  The computer was another.  There have been countless inventions since the shovel that reduced the number of people it took to perform some kind of work, and each time we reduce the number of people necessary to produce a given level of output, many in the public panic about what life will look like once work becomes obsolete.

In a similar light, I used to get Jehovah’s Witnesses at my door all the time.  One day I let them in and made coffee.  As we talked, I asked them if Jehovah’s Witness was something new, or a truly established religion.  “Oh – it’s quite established,” they assured me, “going all the way back into the 1870s, with the teachings of Minister Charles Taze Russell!”  This led to a discussion about how Jehovah’s Witness differs from other Christian-based theologies.  As we went over how nobody knows the precise time Christ will return, I eventually got out of them that Jehovah’s Witness teaches that the end is right around the corner, or ‘The End is Nigh,’ as people sometimes say.  At this point, I asked them if I am really to understand that Jehovah’s Witnesses have been teaching that the end is upon us, since the 1870s?  “I would have thought that after the first hundred or so odd years, you might have opened up to the possibility that you have it wrong,” I said.

The Jehovah’s Witnesses left, and I have not heard from them since.

How many times must we hear that ‘the end of work is nigh’ before we realize that there is no end to work?  Today’s most powerful computers still operate on software, and are still only capable of working within their programming.  Programming has gotten vastly more complicated, and our ability to simulate intelligence has been vastly improved, but computers are no more intelligent in any real sense – no more self aware and no more capable of actual thought – than was the first Texas Instruments calculator, from the 1960s.  As long as there are things, like thinking, that computers are not capable of doing, then while computers and robotics will take an ever larger chunk of mundane and repetitive jobs from us, computers and robots will only enable people to do more, and it is time for these hysterical fears of ‘an end to work’ to go away.

Work will never be obsolete.

Shoveling with one’s bare hands is obsolete.  Shoveling large amounts of dirt with a shovel is obsolete.  Making buggy whips (for horse-drawn carriages) is obsolete.  The abacus is obsolete.  Memorizing large numbers of legal precedents is about to become obsolete, as is the ability to memorize symptoms and causes in healthcare.  Will this eliminate the need for lawyers and doctors, or will this only make them more effective?  If a doctor can see ten times as many patients with the use of computerized diagnostics, and if computers and robotics can assist with things like surgery, it will lead to both better, and cheaper, healthcare.  If you need heart surgery, do you want a surgeon to run a zip-line across your chest, and operate with a scalpel, or would you prefer the precision a surgeon can achieve using a computer and robot to do the actual cutting, with as little cutting as possible?  I for one would prefer surgeons to work on continuously making the computers and robots better, than having the surgeon holding the scalpel.

If surgeons lead computers and robots to do surgery, will that lead to the need for fewer surgeons, or will it make a whole myriad of new kinds of surgery – things that require far more precision than is possible today – possible?  The answer is what it has always been with automation: robotics and AI will lead to more, better, and cheaper goods and services, but not fewer jobs.

We will need fewer jobs to produce the same output we produce today, but the simple fact is that in the future, we will not produce the same output we do today.  We know intuitively that the pie does not have to stay the same size when we talk about economics.  Why do we revert to socialist-like thinking when we talk about automation?  In the future we will produce more of some things, we will produce fewer of other things, and we will produce many things that have not even been invented yet.

Automation does not replace people so much as it moves them from activities that add less value, to activities that add more value, and there are an endless number of ways people can add value to the world around them.  Instead of working within a production line, people may maintain that line.  Other people may work on systemic changes to improve the production line, or may work on product changes to improve the products being built.  Automation also has a ton of drawbacks, such as that a fully automated line can only look for defects that it knows to look for.  When a new defect occurs – something the system is not designed to detect – the number of products that may have been made defective before the defect was found (and the cost of fixing the defected items) can be staggering.  People, on the other hand, are intuitive in looking for problems, and are much more apt to see a new problem arise.

Another problem with automation is that it retards progress.  Ford almost went bankrupt in the 1980s, when Japanese automakers began introducing front-wheel drive cars and cars with electro-static painting systems (that bond the paint to the metal on a molecular level).  Ford looked at the cost of retooling it’s production systems, which were highly automated, to implement these improvements, and they found that it would have been prohibitively expensive, so they did not implement those technological improvements until lost market share forced them to.  Imagine trying to run a company that cannot afford to keep up with its competitors.  Does that sound like ‘improvement’ to you?

The future is about improvement rather than automation, and automation will only work when and where it improves operations.

I once toured a manufacturing facility that had recently converted the creation of a particular sub-assembly, from a time-consuming manual process, to a fully-automated robotic process.  It was very impressive, watching robots effortlessly put together a complicated sub-assembly, one after the other, without people.  I mentioned that the automated process must have been expensive, and was told that while it was expensive, as long as they run it at 80% efficiency, or higher, it costs less than did the manual process.

I noticed that forklifts were taking pallets of this particular sub-component to trucks.  I asked why, and was told that while the automatic process could build the sub-component much faster than could the manual process, the company could not use all of the parts being built, and was sending them out to be stored off-site.  I also noticed that much of the plant was devoted to feeding this automated process, to keep it always running at 80% efficiency, or higher.

In reality, a company does not make money building sub-components.  Companies make money building (and selling) finished goods, and it is the throughput of the overall organization, in terms of the goods and services it can sell into the market, that is important.  Inventory may be listed as an ‘asset’, but it is hard to make money building ‘assets’ faster than they can be sold.

What companies need is balance between supply and demand.  As demand grows, the company must be able to ramp-up production to meet that demand.  It is equally important for a company to be able to scale-back production when demand falters, and it is interesting to note the large number of former-great companies that grew like gang-busters while the demand for their products grew, and then collapsed at their peak.  General Motors was the largest corporation in the world when it produced my 1978 Corvette, but a a few years later it only existed because of government assistance – something it would not have had in a true free market.

Manual processes scale up just as well as they scale down.  Automated processes do not.  This observation does not make automation ‘bad,’ but it does mean that automation should be targeted toward certain types of activities, such as those that are dirty, demeaning, dangerous, or difficult.  Processes are also usually better when the only parts that are automated are the parts that are dirty, demeaning, dangerous, or difficult, and when those parts are automated only to the point that they are not dirty, demeaning, dangerous, or difficult.  Beyond that, companies need people, as people are far more adaptable than are robots, making us far better at handling, and even creating, change.

Dr. Mark Perry, who was one of my economics professors at the University of Michigan, has done a lot of detailed research on the effects of automation on the lifestyles of workers.  Do you know what he has found?  It turns out that market dynamics (like competition) create pressures that correlate the price of goods and services with the number of human work-hours it takes to produce them.  The more human-hours it takes to produce something, the more hours people have to work to be able to afford that thing.  The evidence for this phenomenon is overwhelming, and it goes back, as I said earlier, to the invention of the shovel.  How many hundreds of thousands of years will it take before we learn that the shovel is our friend?

I cannot tell you what the future looks like with any precision, but I can tell you that as long as we allow the markets to decide, it will be brighter than today.  You will still have a job in the future, and you will add more value to the world than you do today.  While many people will lose their jobs, there will always be other jobs, and our lifestyles are only going to improve over time.

Stop fretting about a future without jobs, and when you see someone else flipping out about it, point them here.  We have absolutely nothing, in the long run at least, to be afraid of, so stop worrying and enjoy being alive.

2 thoughts on “The AI Unemployment Myth”

  1. I highly recommend you read Sapiens: A brief history of humankind by Yuval Noah Harari with particular attention to the last three or four chapters.

  2. Automobiles were my world for 25 years, and these are my thoughts. In the 70s, Japanese were making small cars bc they had high gas prices, the people and country were small in stature and area, and they had less access to steel. In the US, we favored the big car; the only salable small cars were sport or performance cars. Pintos, Mavericks, Astres, Corvairs, or Chevettes were hardly worth it bc they cost almost as much to build as Mustangs, T-Birds, or Camaros, yet had to be sold for so much less. Decisions were based on demand and forecasting.
    The gas lines of the 70s caught the US unprepared – and our development time was longer than most imports. Forecasting failed in the 70s. So Datsun and Toyota were ready with their subcompacts and made huge inroads. Ford and GM introduced assorted small cars, but all the big 3 partnered with an import company. Chrysler had a 25% share in Mitsubishi, and sold the Colt and Ram 50.
    Chrysler developed the first sub-compact FWD car, the 1978 Omni/Horizon, but it still needed a government loan. After the 1981 Aries/Reliant (FWD compact) and the breakthrough 84 Caravan/Voyager, the loan was paid back. I don’t recall GM getting assistance.
    While we were working on small cars to meet demand, Japan was using their subcompacts as stepping stones, and they started working on larger and more luxurious cars to capture the US market. With their new plants that the US helped build and with Japanese government subsidies, they were smart enough to wage economic war on the US.

Leave a Reply