(This is a 3-part series, link to next post is at the end)
Here’s a new factory for making automobile frames, specifically designed to minimize the need for human labor. The CEO of the company that built it actually said, “We set out to build automobile frames without people.”
At the start of the process, rough steel plates are inspected by electronic sensors, automatically pushing aside any that deviate from tolerances. Conveyors take the plates through punching, pressing, assembling, and nailing machines, as well as a machine that can insert 60 rivets simultaneously in each frame. A set of finishing machines then rinse, dry, spray-paint, and cool the frames. Aside from a few men moving frames between conveyor belts, the floor routine of the plant requires almost no hand labor.
And today’s robotics and artificial-intelligence advances go far beyond automating routine manufacturing labor and take over the kind of cognitive functions once thought to be exclusive to human beings. Here, for example, is a new AI-based system that displaces much of the thought-work which has been required of the people operating railway switch and signal installations:
The NX control machine is in effect the “brain” of the system. It automatically selects the best optional route if the preferred route is occupied. It will allow no conflicting routes to be set up. It eliminates individual lever control of each switch and signal.
Pretty scary from the standpoint of maintaining anything like full employment, don’t you think?
Actually, the first passage is a description of the A.O. Smith frame factory which was built in…1922. The second passage is taken from a brochure by General Railway Signal, dated 1954. (Although this technology was originally introduced by GRS in 1936.)
I’m pretty confident that if the above passages were to be turned into press releases and sent to the business and general media, then there would be dozens of stories using them as an example of the way in which robotization is destroying jobs.
Articles on the job elimination expected as a result of robots and artificial intelligencesand there seem to be dozens of such articles coming out every week—very rarely include any attempts to put current innovations in historical perspective. They also rarely include much in the way of meaningful numbersI searched through quite a few articles on a new automated shoe-making factor in Germany, for example, without getting any light on what is clearly a very key metric: the ratio of shoes produced to number of people working in the factory. The answer to the question of whether what we are seeing now is really a sharp upward break in the labor-productivity line…or, alternatively, only a continuation of long-standing trends….seems usually to be simply implicitly asserted, rather than researched and defended.
Certainly, the recent labor-productivity statistics (also this) don’t seem to provide much support for the idea of a sharp upward productivity breakalthough this may be to some degree a factor of the lackluster economy.
In the next post of this series, I’ll attempt to provide some historical perspective on today’s automation technologies by sketching out some of the past innovations in the mechanization of work, focusing on “robots,” broadly-defined…ie, on technologies which to some degree involve the replacement or augmentation of human mind, eye, and hand, rather than those that are primarily concerned with the replacement of human and animal muscular energy.
Note: the description of the A O Smith auto frame factory is taken from the book Inventing Ourselves Out of Jobs?, by Professor Amy Sue Bix, which will also be extensively cited in the next post of this series.
Someone once said that progress consists of replacing problems with more difficult problems.
Go back not that far into history, and everyone (from young children to the doddering old) had to work all day, every day just to have a shot at having clothes on his back, a roof over his head, and food in his stomach. And many times in history, no matter how hard individuals worked, they were still cold & hungry.
Mechanization allied to liberal use of fossil fuels has solved those problems, and people in the developed world are generally adequately clothed, housed, and fed. But now we face the problem of perhaps not being able to find meaningful work for everyone who wants the dignity of labor or the challenge of being productive.
There can be a certain parallelism between those who say not to worry about further automation and those who say not to worry about cheap imports. Somehow or other, the “invisible hand” will take care of it. In the case of the free traders, they are clinging to this delusion even though the evidence of their own eyes shows that workers displaced by cheap imports do not automatically end up in better paying, more satisfying work. International trade has brought us great benefits in the past, but the social costs of international trade are clearly rising; we may already have passed the point of overall diminishing returns.
It is possible that further automation make create a similar issue of (social) costs rising faster than (productivity) benefits. In which case, as societies we will have a very difficult problem to address.
Great series of posts, David. You are singing my song.
I would add two things:
The best definition of quality, for manufactured goods and many services is “absence of variation” conformance to specification is another, more common, definition but I don’t think it is as useful.
It is quality that gives us cars that routinely last 2-300,000 miles, for example.
Automated processes depend on quality parts. But the only way to get quality parts is by automation. Any human touch will degrade quality.
I don’t think that the improved usability of everything we use every day gets enough recognition.
2) there seems to be a great deal of confusion between loss of manufacturing jobs and loss of manufacturing.
We have lost a lot of manufacturing jobs. A few years ago I pulled the numbers on manufacturing output going back to @1950 from the statisical abstract. I adjusted for inflation and population to get per capita output in constant dollars.
There were only 2 times it didn’t increase year over year. Here are 2 data points:
1970 $1,225
2010 $6,011
In 40 years manufacturing out per capita increased more than 5 times. Not too shabby.
John Henry
Lest anyone think i believe it possible to achieve “absence of variation”, I don’t. All processes will vary some and all outputs will vary some.
It is more of a philosophy than an actual goal. Among other things it recognizes that any variation imposes a quality cost. An acceptable cost, if in specification, but a cost nonetheless.
The philosophy emphasizes thst just being in spec is not enough. The focus must be on driving out variation.
Bing “taguchi loss function” for more info. Wikipedia has a good article.
John Henry
David, the Bix book looks fascinating. I was disappointed to see that there is no kindle version. I really hate it when authors do that.
Since my son first got me a kindle6-7 years ago, I no longer have the patience for paper books.
I wonder if you have ever read Henry Ford’s books? Particularly his first “my life and work” 1922.
Best book ever on manufacturing in general and the Toyota Production System/lean manufacturing in particular.
Out of print in English for 75 years. Never out of print in Japanese.
I have electronic versions of all 3 and am happy to share. Just drop me a note at johnfajardohenry@gmail.com
Not just David but anyone interested.
John Henry
Gavin…”There can be a certain parallelism between those who say not to worry about further automation and those who say not to worry about cheap imports. ”
I question whether *automation* is equivalent to *offshoring* in terms of its impact on jobs and on worker well-being.
As a thought experiment, what if Henry Ford had, instead of introducing the assembly line and other productivity improvements, “offshored” the Model T production to workers earning 10 cents a day in Mexico?…say he had been able to meet the Model T cost and price targets without any need for assembly lines and such, just raw cheap labor.
Would the impact on American prosperity have been equivalent to that which the Model T, made in America, actually had?
John Henry…have not read Ford’s book, but I did read Sorensen’s ‘My 40 Years with Ford.’
I see the Ford book is now available on Kindle for 99 cents.
Referring to the palletizing robot:
The simple pallet, no moving parts, is a pretty serious alternative to labor, even without robots.
If, for instance, transportation costs are low and labor costs are high, the cheap way to ship a load is to palletize once at shipping end, pay for the pallet and the weight and cube it consumes in the transport, and take the whole unit load off at the receiving end. If transport costs are high and labor costs low, then the unit load is broken down and depalletized before loading by shipping laborers, the goods are shipped high and tight without the dunnage weight of the pallet, and the receiving laborers then build up their own pallet / unit loads as they off load.
Standard sized pallets exchanged in pallet pools — CHEP pallets in the US and EURO pallets dating back to the Common Market — also reduce a lot of laborious handling. A reused pallet being cheaper than a one-time-use pallet, they’re used more, and so more large units are handled with forklifts and other MHE than previously handled, er, by hand, one carton or bag at a time.
Robots are fun to envision and talk about but invisible, silent, 2 labor-reducing technologies are all around us.
David F: “As a thought experiment, what if Henry Ford had, instead of introducing the assembly line and other productivity improvements, “offshored” the Model T production to workers earning 10 cents a day in Mexico?”
David — that is a very interesting thought experiment. I am not well-informed about Henry Ford’s history — but I do recall reading somewhere that one of his aims was to get to a condition where the guys who built the Model T could afford to buy it. Raising wages in the auto plants through higher productivity may have had a knock-on effect of raising wages elsewhere in the economy, in a beneficial cycle. Part of the overall economic benefits from the Model T came from manufacturing it in the US, and part from using it to improve transportation within the US. I have no idea what that split was — 90/10? 50/50? 5/95?
Whatever that split of benefits may have been, if the Model T had been imported from Mexico, and the workers in the US had not become more productive through “Fordism”, it seems likely that a substantial part of the economic benefits of improved mobility in the US would have been captured by Mexico. The US would have had to produce & export something else to Mexico in exchange for those Model Ts.
It seems self-evident — over the long term, any economy cannot consume more than it produces.
Certainly one major reason for the $5/day wage was to reduce what had been the very high turnover level among workers. A lot of people really didn’t like the work. In early 1914, a worker’s wife wrote (anonymously) to Ford as follows:
“The chain system you have is a slave driver! My God! Mr Ford. My husband has come home & thrown himself down and won’t eat his supper–so done out! Can’t it be remedied?”¦That $5 a day is a blessing–a bigger one than you know, but oh they earn it.”
…the implication being that they *wouldn’t* have put up with the assembly line work without the high wages.
I believe Ford also hoped that others would adopt the high-productivity high-wage approach, thereby increasing his potential market (clearly, Model T purchases by Ford workers themselves would not have been enough to drive the product’s success!) and maybe he also cared at some level about worker well-being. But fundamentally, I bet he wouldn’t have done it if he didn’t have to.
Pouncer: You are right. Shipping containers are simply an extension of the concept. I would extend your preconditions to also include roads that can deal with trucks and common availability of things like pallet jacks and forklifts. This is basically describing the first world.
When we have to ship something particularly sensitive, we strap it down to a pallet and ship by freight because we’ve found that UPS takes a fragile sticker as a challenge.
I often have to assemble something from a lot of different parts. I’m always a little surprised how long the process I call shucking takes. This is where I remove the various components from whatever packaging they arrive in.
A story my father told is how the transmissions for the Model-T arrived at the factory in a box that was broken down to form the floor boards. Now parts arrive with the absolute minimum of packaging, usually on racks ready to roll directly into production. This extends to things like electrical receptacles that arrive at building sites in large bulk containers with the screws backed out so that they can be installed more quickly.
I worked at a plant that produced pre-stressed concrete items that was significantly automated. We had a plant in Columbia where we were forbidden from even a lot of common labor saving machinery. The quality was poor and the plant had been shut down before I started.
Look at how much trouble Amazon has with trying to package individual orders for shipment. They still have trouble getting items to the purchaser in one piece and avoiding the box in a box in a box.
Look at how much trouble Amazon has with trying to package individual orders for shipment. They still have trouble getting items to the purchaser in one piece and avoiding the box in a box in a box.?
I ordered a $650.00 pool pump in July. My pool guys wanted almost a thousand for the same pump.
The pump arrived, poorly packed and , when my plumber started to install it, found it cracked. Probably dropped.
I returned it and the new one arrived, no better packaged, but intact. Fortunately, Amazon’s return service is always good.
It was interesting to see how poorly packed an item of that value was shipped.
You would think Amazon would have built enough sense into their system, something like: if weight > 30# then special packaging required. If it was from China, I bet it was in a foam packing surrounded by very poor quality thin fiber board.
All it would take would be a little judgement to see that the box wouldn’t stand up. You’re lucky two ways: First, if the box allowed you to return it without the hassle of packing it yourself. Second, that you didn’t order something else at the same time that was packed loose in a big box and crushed.
This is why I have trouble believing all the hype about AI. When are we going to see some evidence that it will actually solve problems. How long before I can type 3/8″ HHCS and find actual hex head cap screws instead of a random list of items with 3 and 8 and H somewhere in the description. Damaged returns are a big cost to Amazon, the progress they are making isn’t very apparent. They are, of course, keen to sell you another pump I bet.
Great comments about pallets, Pouncer. Here’s a good article about pallets in general and CHEP in particular http://www.cabinetmagazine.org/issues/52/hodes.php
Pallets are another one of those deceptively simple things that have had enormous impact.
I tend to disagree with you about loading and unloading pallets if labor costs are low. Pallets add little to the shipping cost in most cases. Palletizing in the factory and not depalletizing until final destination seems like it will always be cheaper.
1) The pack patterns will be optimized at the producer, they may not be if repalletized at other pints in the supply chain.
2) Any time the individual product is touched, there is a possibility of damage or pilferage. If loading 1,000 cases of Scotch on a truck, it would be easy to pull a couple bottles from an individual case. Nearly impossible to pull it fromn a stretch wrapped pallet.
3) The unitized pallet load, especially with edge reinforcements added under the wrap, will be much less subject to damage in transit than individual cases.
4) Inventory control is easier since you only have to count perhaps a dozen pallets instead of 250 cases.
Not to say there will not be some instances where it might make sense, just that these would be fairly rare exceptions, I think.
Pharmaceutical plants in general, as well as some food and other plants, do not permit wooden pallets in the plant. This is because of known and potential contamination. Raw materials come in on wooden pallets, because nobody can afford a float of plastic pallets. In the warehouse, the pallet load is shifted to a plastic pallet for storage and use in the plant. Finished goods are palletized on plastic pallets and in the warehouse, before shipment, are transferred to a wooden pallet.
Transfer is done with a machine that basically pushes the entire pallet load from one pallet to another. Individual cases are not handled until used.
The shipping container seems pretty mundane and a no brainer. Most people don’t realize how relatively new it is. They only really came into use in the 1970s. Shipping containers required, not just the container itself, which is a pretty ingenious invention it also required the invention of:
1) Ships
2)Trailers
3) Ports
4) Labor force
5) Documentation
6) and more…
And nobody wanted it other than a chicken hauler from North Carolina named Malcolm McLean. The story of “The Box” is pretty amazing. The book The Box by Marc Levinson is a great read. (If anyone remembers A&P, which once sold more than half of all the groceries in the US, he wrote an excellent history of them too.}
John Henry
David Foster Says:
I believe Ford also hoped that others would adopt the high-productivity high-wage approach, thereby increasing his potential market (clearly, Model T purchases by Ford workers themselves would not have been enough to drive the product’s success!) and maybe he also cared at some level about worker well-being. But fundamentally, I bet he wouldn’t have done it if he didn’t have to.
He did hope that others would adopt his methods and wrote and spoke about them constantly. He was continually amazed that others didn’t. Toyota (as Toyoda) had been in the textile machine business when they decided to build cars. Henry Ford welcomed them into the plant and showed them everything. As he did to anyone he could get to listen to him. He laid out all the details (including a chapter on the $5 wage) in “My Life and Work”
“The payment of five dollars a day for an eight-hour day was one of the finest cost-cutting moves we ever made, and the six-dollar-a-day wage is cheaper than the five. How far this will go we do not know.”
He complained long loud and publicly that other manufacturers didn’t follow him. I wonder what would have happened if they had? The $5 was like golden handcuffs. Men (and women) could not leave to go elsewhere without experiencing a dramatic pay cut. If others had been competing for the same workforce, the $5 might not have been the competitive advantage that it was.
Some have claimed that he raised the wage so his workers could afford to buy cars. Sounds nice and that was an effect but was not the reason. Ford’s vision statement with the Model T was:
I will build a motor car for the great multitude. It will be large enough for the family but small enough for the individual to run and care for. It will be constructed of the best materials, by the best men to be hired, after the simplest designs that modern engineering can devise. But it will be so low in price that no man making a good salary will be unable to own one – and enjoy with his family the blessing of hours of pleasure in God’s great open spaces. – Forward to My Life and Work.
He was relentless about driving down costs. The first Model Ts cost $800 or so at a time when all other cars cost $5000 or more. By 1927 when it ended its run, they cost under $300.
His workers, and other workers, were able to afford Model T’s not because he paid them well but because he had driven the cost down. Part of the cost reduction was the increased pay. Counter-intuitive as that may sound.
John Henry
Part of the reason for the $5 wage was out of the goodness of Ford’s heart, so to speak. It was not technically a $5 wage. The wage was actually $2.50 which was more or less the going rate. Ford believed that he was getting too much profit and it should be shared with the workers. Today profit sharing is mainly deferred as a year end bonus, 401K or the like. He thought they should get it in real time so paid it in the wage packet.
It was effectively a $5 wage until 1928 when profits fell and he tried to reduce the profit sharing.
Some other things Ford did for his workers:
The $5 wage caused stores in the areas where his workers lived to raise their prices gouging his workers. He opened his own stores, for his workers, selling at normal retail prices. This helped drive down prices.
Medical care was a problem and when he was asked to subscribe to a fund for a modern hospital, he found the borad not doing a good job. He refunded all subscriptions and completed and operated the hospital with his own money. Management was along the lines he managed FoMoCo.
He found that his workers were not spending their money wisely, in his opinion, and formed a social department. Social workers would go into homes and help organize them, teach budgeting, cooking and so on. Turned out to be too meddlesome for most people and he abandoned it in about 2 years. Still, his heart was in the right place. And it is not much different from what today’s govt social workers do to the people on their “payroll”
Henry Ford was a fierce environmentalist. He makes AOC and her Green New Deal look like a smokestack baron. Lots of detail on his various environmental initiatives at Ford in his 3 books but especially in the 2nd, Today and tomorrow.
I am a huge fan of Henry Ford as an entrepreneur, industrialist, manager, visionary. The man himself seems like something of a crank, though. I think I would have enjoyed working for him. Not sure I would have enjoyed hanging out with him.
David mentioned Sorenson’s book. He was, essentially, general manager of Ford Motor Company for 40 years. He is the one who implemented many of Ford’s visions. Some of Ford’s visions, the moving assembly line, are probably more Sorensons than Fords. I highly recommend that book too.
John Henry
But now we face the problem of perhaps not being able to find meaningful work for everyone who wants the dignity of labor or the challenge of being productive.
This is because everyone wants to be given a job that is easy and not much actual work.
For example, if our goal was to provide the dignity of labor to everyone, then learning a classical vocation (e.g., woodworking or welding) which allows the input of artistry into the process is a clear option. Most people can learn to do woodworking such that they can “value add” to a cabinet or other a stylistic design which makes the cabinet unique and not just another piece of ‘x’ from a factory where 999 other identical pieces were made.
Of course, woodworking is a matter for experience and practice, and skill in it is not something trivially easy to come by. It takes perseverance, patience, and determination to learn to do well.
The same thing with welding. Making wrought iron into fanciful shapes is another thing requiring practice and effort to learn the craft.
For both, the resulting ART produced can be quite fulfilling, and should be encouraged as an option — a highly respected one — for the future.
“For both, the resulting ART produced can be quite fulfilling, and should be encouraged as an option ”” a highly respected one ”” for the future.”
I wonder what the total current size of the handicraft market…individually-made furniture, art-objects, paintings–might be? Haven’t seen it cited anywhere.
Whatever that split of benefits may have been, if the Model T had been imported from Mexico, and the workers in the US had not become more productive through “Fordism”, it seems likely that a substantial part of the economic benefits of improved mobility in the US would have been captured by Mexico. The US would have had to produce & export something else to Mexico in exchange for those Model Ts.
One of the issues this thought experiment ignores is that the quality of the American worker at the time was probably considerably higher than that of the typical Mexican worker. They had a different work ethos, a different educational level, and so forth, all leading to a much higher quality of product from a production line consisting of those workers.
Here’s the Amazon link to The Box, by Levinson:
https://www.amazon.com/gp/product/0691136408
I’m reading it now. Fairly good.
For most of the Model T era, northern Mexico was a war zone.
John Henry Says:
September 3rd, 2019 at 12:34 am
Lest anyone think i believe it possible to achieve “absence of variation”, I don’t. All processes will vary some and all outputs will vary some.
It is more of a philosophy than an actual goal. Among other things it recognizes that any variation imposes a quality cost. An acceptable cost, if in specification, but a cost nonetheless.
The philosophy emphasizes thst just being in spec is not enough. The focus must be on driving out variation.
Bing “taguchi loss function” for more info. Wikipedia has a good article.
John Henry
End Quote.
Years ago while traveling, I had a conversation with my seatmate about product quality. He happened to work for Cummins.
I told him of an idea I had about product tolerances. Each successfully produced part will be within the specified tolerance. If the parts are measured and given an identifiable tag, their measurement can also be kept available. When parts are assembled, the combined product items, such as crankshaft and connecting rod, can be assembled from those whose tolerance measurment combination insures the ‘best’ possible product.
My father spoke of ‘stackup of tolerances’ leading to inferior or unusable product assemblies. Using the computerized measurement system already used in production in combination with a database of those measurements, and a tracking system, could produce product that was overall closer in tolerance than the tolerance specified. You would have better quality product with nothing additional except some tracking and sorting. At least I thought so 25 or so years ago when I proposed such a system to a diesel engine producer. I have no idea if that idea was ever implemented. Of course if product had no variance, such a system would be a waste of time.
tom
“When parts are assembled, the combined product items, such as crankshaft and connecting rod, can be assembled from those whose tolerance measurment combination insures the ‘best’ possible product.”
I believe this is sort of like what was done by the craftsmen assembling products in the pre-mass-production era…although they didn’t do it via measurements, it was more like which combination of parts actually *fits*.
Interesting idea to revive the approach using modern measurement & tracking methods…not sure if the variations in most cases are sufficiently large to justify the additional labor, though.
There are quite a few items that are produced this way. It’s usually done by separating the parts into different bins according to the significant characteristic.
As an example: Hydraulic valves have a spool and body. The only way that they can be sealed is by making the gaps between the spool and body as small as possible, when if they are too small, the valve will bind or seize. The parts ground and lapped to the closest tolerance possible will be individually measured and put into actual bins according to the very small difference in diameter. Assembly then consists of choosing a spool to match a body. This is often expanded to consider something like operating temperature where a match that would work at one temperature will not at either a higher or lower temperature. Such valves usually must either be replaced as a unit, A part matching the binning stamped on it procured or a universal replacement with degraded performance used.
To expand this to an entire engine would maybe be possible in the factory and is in some limited cases for some parts. The problem comes from repairs. This greatly increases the parts inventory requirements and heavy duty engines are repaired a lot. Every special matched part increases the chance that a mistake will be made. Most of the progress that John Henry talks about has been accomplished by a combination of reducing production tolerances and making design improvements to accommodate achievable tolerances.
I told him of an idea I had about product tolerances.
You were ahead of your time Tom. Statistical tolerance analysis is now used in industries with large assemblies like automotive and aerospace. CETOL by Sigmetrix is one of the big platforms, used in combination with production line and process plan studies, to generate data that’s stored and indexed into process capability libraries.
Besides dimensional tolerance stackups, geometrical tolerances can be analyzed by Monte Carlo simulations and integrated with 3D metrology systems. Although, that’s somewhat limited to part orientation and arrangement. The holy grail would be to scan any part in any position just prior to assembly, instantaneously compare it to the 3D model, then accept or reject it in-process. But that’s another one that’s probably more like the proverbial dogged chase after the speeding bus of perpetual potential.