How Apple learned automation can't match human skill

Posted:
in General Discussion edited June 2020
Apple has spent years and millions of dollars on automating its production lines with technology, and it has always reverted to using skilled human beings instead.




The most profitable technology company in the world, and arguably the most technologically advanced firm in history, won't use automation to make its products. Apple has repeatedly tried to build machines to build its machines, but in every case bar its recycling plans, it has failed -- and reverted to using human beings instead of robots.

"Robotics and automation is fantastic and amazing when it works," David Bourne told The Information when discussing his time working with Apple supplier Foxconn. "But when something breaks, God knows what happens."

Bourne, who is now principle systems scientist at the Robotics Institute at Carnegie Mellon University, used to work with Foxconn on multiple automation projects that had been intended to produce Apple's products.

It was Foxconn that pressed Apple on automation, and it was also Foxconn that impressed Apple on the idea. In 2012, Apple executives including Tim Cook were reportedly in China to see the results of an experimental production line. It was built to assemble the iPad by robots and showed the parts being cut, polished and then partially assembled into the final product.

Terry Guo, then chairman of Foxconn, reportedly told Apple that his firm's assembly lines would contain one million robots within two years.

Seven years on, in 2019, Foxconn was using just 100,000 robots across all of its manufacturing. Neither Foxconn nor Apple would comment publicly about why the automation was so much lower than predicted, but according to The Information, sources say it's down to dissatisfaction by Apple.

Apple's own automation teams

Seemingly as a result of Foxconn's original efforts, Apple launched its own secret robotics lab in 2012, based around six miles from Apple Park. It housed a team of automation specialists and robotics engineers who initially tried mimicking the iPad automatic production line.

The iPad Air - used by hands, made by hands.
The iPad Air - used by hands, made by hands.


They were given the target of reducing the amount of human labor needed by half. Specifically, Apple wanted to be able to cut 15,000 workers from the production line, which represents about 50% of the number of workers used at key times.

It didn't work. Typical problems that arose include how Apple's use of glue required precision the machinery couldn't reliably match. And the tiny screws needed required the automation to correctly pick and position them but that same automation couldn't detect problems the way a human hand could.

This lab was abandoned in 2018, although reportedly some of its work was picked up and continued by other parts of Apple. It wasn't the only department working on the project -- and arguably wasn't the biggest failure.

That title goes to the millions of dollars spent automating production of what would become the MacBook in 2015. That automated production line was started the year before in 2014, but persistent failures meant not only was it also abandoned, but that MacBook itself was postponed by some months.

MacBook

The attempt to automate the production of the MacBook went beyond a test department in Cupertino. The equipment was installed in a factory in China, and it was intended to assemble the screen, the keyboard, and the trackpad into the MacBook's casing.

Reportedly, though, there were problems with even the conveyor belt that moved parts along the line. It was erratic, it was sometimes slow, but the greatest issue was that parts along the line kept breaking down.

The MacBook was delayed to 2015 because of automation failures
The MacBook was delayed to 2015 because of automation failures


What was worse, it wouldn't always be clear that something had gone wrong. "If things stop working, the automation can't detect that all the time and repair it," Bourne told The Information.

Insurmountable issues

If any firm were capable of fixing a technology problem, it would surely be Apple, but alongside technical issues there were more fundamental ones. Most specifically, since Apple redesigns its major hardware in at least some way every year, it would also have to redesign the automated factory lines.

Compared to that, training workers on new designs is vastly easier and quicker.

Plus, one reason both Foxconn and Apple were interested in automation is that as well as being dependent on workers, that dependency fluctuates greatly. Foxconn was having problems recruiting enough staff for the peak periods just after, say, an iPhone launch.

Automation would theoretically reduce that problem, but Apple already has a way of removing it. When it needs to, Apple is currently able to switch production to other companies. If there's a problem or if more production is needed, it has alternative sources it can leverage.

If those sources had to have annually-updated automated production lines installed by Apple, the company would not have the flexibility it has now.

So more than overcoming technical issues, it is the series of business rationales that are likely to keep Apple using human labor in the production of its machines. It's not just Apple, either, as both Tesla and Boeing have famously attempted and abandoned automation for the same reasons.

Comments

  • Reply 1 of 14
    hammeroftruthhammeroftruth Posts: 1,051member
    So Skynet would need humans to make terminators. It just goes to show you that nothing can replace humans.... yet
    JWSCOferfastasleepthtwatto_cobra
  • Reply 2 of 14
    lkrupplkrupp Posts: 9,301member
    Bottom line? Don't expect iPhone/iPad/iMac/MacBook manufacturing to come to the U.S. any time soon. It will always be a pipe dream.
    lolliverDAalsethcanukstormGeorgeBMacwatto_cobra
  • Reply 3 of 14
    This article is slightly misleading. Much of what goes into producing an Apple device is heavily automated. Final assembly on the other hand is usually pretty manual. That is one of the things which drives component consolidation and design simplification. Despite have a lot more features the number of parts in the iPhone has stayed pretty much the same.
    neilmwatto_cobra
  • Reply 4 of 14
    sunman42sunman42 Posts: 127member
    lkrupp said:
    Bottom line? Don't expect iPhone/iPad/iMac/MacBook manufacturing to come to the U.S. any time soon. It will always be a pipe dream.
    As Steve Jobs explained to President Obama in 2011, it's not the labor coasts that drive where Apple manufactures its products, but the ready supply of manufacturing process engineers and access to international supply chains. I believe I recall Mr. Jobs's saying that the difference in labor costs per iPhone would be in the single digits of dollars if they were assembled in the US. We just don't have the supply of (associate's degree-level) process engineers. Makes sense, as it's only in the final assembly that any human is involved in the process.

    Frankly, this story suggests to me that (1) our robotics engineers aren't that clever and (2) industrial robotics has been focused on large system assembly (say, cars), rather than fine object assembly. New technologies will have to be researched and implemented to make the latter possible; the only question is whether it makes sense financially.

    Remember that since 2008, Mac Pros have been assembled in the US, in a largely automated plant — though fine assembly is still done by humans. I suspect the labor involved in that manufacturing process is also a negligible cost factor.... especially considering the denominator. 
    edited June 2020 OferfastasleepGeorgeBMacwatto_cobra
  • Reply 5 of 14
    seanismorrisseanismorris Posts: 1,624member
    If you can build a robot to recycle an iPhone, you can create a robot to build one.

    The question is it cost effective or the most efficient solution?  The answer today is no, but it’s only a matter of time before the answer is yes.

    We see this everywhere.  You start with a pipe dream, you build a prototype, everyone gets overly excited, then reality brings things back to earth.  But, eventually some stubborn SOB gets it to work...

    We’re seeing the same thing in AR/VR.  Right now, most pocket books are empty and most are abandoning the idea as to hard.  But, is there any doubt it’s coming?

    Robots replacing skilled labor is also coming.  It might take another decade, but there’s no doubt in my mind it’s going to change the world.

    ....
    In the last 20 years, there’s been incredible progress in robotics.
    https://www.cnn.com/videos/business/2020/05/11/spot-coronavirus-robot-singapore-orig.cnn-business
    edited June 2020 sconosciuto
  • Reply 6 of 14
    fastasleepfastasleep Posts: 5,477member
    Have they tried beating the robots when they screw up?
    DogpersoncornchipGeorgeBMacwatto_cobra
  • Reply 7 of 14
    Have they tried beating the robots when they screw up?
    In metropolis they did lol.
    fastasleepwatto_cobra
  • Reply 8 of 14
    dewmedewme Posts: 3,698member
    sunman42 said:
    lkrupp said:
    Bottom line? Don't expect iPhone/iPad/iMac/MacBook manufacturing to come to the U.S. any time soon. It will always be a pipe dream.
    As Steve Jobs explained to President Obama in 2011, it's not the labor coasts that drive where Apple manufactures its products, but the ready supply of manufacturing process engineers and access to international supply chains. I believe I recall Mr. Jobs's saying that the difference in labor costs per iPhone would be in the single digits of dollars if they were assembled in the US. We just don't have the supply of (associate's degree-level) process engineers. Makes sense, as it's only in the final assembly that any human is involved in the process.

    Frankly, this story suggests to me that (1) our robotics engineers aren't that clever and (2) industrial robotics has been focused on large system assembly (say, cars), rather than fine object assembly. New technologies will have to be researched and implemented to make the latter possible; the only question is whether it makes sense financially.

    Remember that since 2008, Mac Pros have been assembled in the US, in a largely automated plant — though fine assembly is still done by humans. I suspect the labor involved in that manufacturing process is also a negligible cost factor.... especially considering the denominator. 
    Great post, although I disagree with the middle paragraph. There’s plenty of expertise in multi axis motion control, I.e., robotics, although the US is a little behind other parts of the world. 

    The term “robotics” is way too generic to describe all of the permutations that actually involve robotics, micro level on up. 

    Size isn’t really a factor at all. Pick and place machines that populate printed circuit boards of every size have been in widespread use for decades and would qualify as robotics done on a micro scale. If you look and MEMS and Microbotics we’re talking Nano scale. 

    Automation, including robotics, is no different than any engineering discipline and practice. Some people and organizations are good at it and some aren’t. Drawing a summary conclusion about such a wide field based on any single implementation, good or bad, doesn’t make sense or paint an accurate picture of the field. 

    You can actually learn a lot about robotics programming by exploring the open source ROS (robot operating system). https://www.ros.org/

    fastasleepwatto_cobra
  • Reply 9 of 14
    XedXed Posts: 975member
    If you can build a robot to recycle an iPhone, you can create a robot to build one.
    You have that backwards. If you can build a machine to assembly an iPhone in its entirety then you can build one to rip it apart, because the latter, destructive step is much easier to accomplish.
    edited June 2020 neilmCloudTalkincornchipwatto_cobrafastasleep
  • Reply 10 of 14
    frantisekfrantisek Posts: 733member
    Just funny quote for weekend :smile: )

    "Robotics and automation is fantastic and amazing when it works," David Bourne told The Information when discussing his time working with Apple supplier Foxconn. "But when something breaks, God knows what happens."


    Yeah. God he knows what happens and loves people so keeps automation process in reasonable pace. And if you do not pray enough you can not get advice from God, that is clear.  :D


  • Reply 11 of 14
    wonkothesanewonkothesane Posts: 1,547member
    I am not sure how many here have some background in industrial engineering. So I will try to explain in simple words. 

    It is not about: can it be done. Rather, it is about: can it be done such that it is sufficiently reliable and at the same time cost effective. 

    It’s simply a difference between tightening 10 screws or ten million ones. 

    We don’t talk about pick and place machines, or dispensing flux:solder paste. 

    We talk very specific processes that imply specific challenges. 

    Now, specifically tightening  on that scale and in this dimensions is this a challenge. Drilling “not so tiny screws” is a very standard process in Industry. And usually the correct amount of tightening is measured through monitoring of torque and/or angle. This works well when the inner resistance of the screwing part in that robot is significantly smaller than what you want to measure -and- the inevitable variation in geometry of the threads as well do not matter compared to what you need to observe in order to monitor your process. With tiny screws as used in the assembly of iPhones etc. this is not state of the art, meaning you can’t buy it off the shelf. 

    There are similar challenges when dispensing glue in that tiny amount as required here. Stuff you do easily when doing this - some skill provided - by hand (such as compensating for change of viscosity based on small variations in ambient temperature, or material age, are sometimes surprisingly hard to implement automatically. 

    It is also no big deal to drill a hole. Drilling 10 thousands in an airplane section, however, is. 

    An experienced machine operator will easily tell you whether the cutting process is running well by simply listening to it. Decades of research have gone into automating this. So far, with very limited success. 

    I am even not surprised about the issues with the ja fling of parts. While it sounds ridiculously easy to move a part from one station to another , the devil is again in the details. Takt time variations can be one, for example, reliably scanning parts, controlled acceleration and deceleration. All depending on the masses you want to move, and of course the required speed. If you go to major companies in automotive or aerospace you will find that handling is very often reduced to the max, or completely eliminated - due to cost and reliability. 

    In all this one has also to consider that one big difference in the consumer electronic industry is that products are in do parison manufactured for a quite short time before a new model comes. This means, that there is simply also much less time for straitening out and issues before the new model arrives. 

    All in all, I am sure that the major subassembly steps, such as making a PCB, are fully automated. It is the more “manual” processes that mount the sub assemblies together (such as screwing or glueing, welding...) are a challenge on that dimensional and production volume scale. 
    dedgeckowatto_cobrafastasleep
  • Reply 12 of 14
    GeorgeBMacGeorgeBMac Posts: 9,862member
    ...

    Insurmountable issues

    If any firm were capable of fixing a technology problem, it would surely be Apple, but alongside technical issues there were more fundamental ones. Most specifically, since Apple redesigns its major hardware in at least some way every year, it would also have to redesign the automated factory lines.

    Compared to that, training workers on new designs is vastly easier and quicker.

    ....

    Yeh, there's always a balance between automation and manual.   And, too often, people assume that automation is always the better of the two.   Not so. 

    My own experience was that almost 40 years ago I was hired as a systems analyst to design and build an automated invoicing system for a start up.  Their Controller reasoned that since it was a technical, IT company it was inexcusable that it was producing monthly invoices manually.  He wanted to "just push a button to get an invoice".   So, they hired me.

    I spent a couple months learning their business and their systems and understanding both.  And, in this case, what I learned was that they produced only a few dozen multi-million dollar invoices each month and, that not only was every invoice unique but most were changed to suit existing conditions and agreements nearly every month.

    I presented the results of my analysis to the Controller and various financial and IT executives and told them:   I can build this for you.  But you won't like it.   It will be expensive (a million $) to build and expensive to maintain (a million $ per year) and it will be clunky and inflexible.   So, my advice was to hirer a spreadsheet specialist who could build and maintain each invoice as needed.   All through that presentation I assumed I had just gotten myself fired by doing away with the main job I was hired to do!   But, instead of firing me (I worked there for another 20 years) they took my advice.  But it took the Controller over 10 years to forgive me.   But, he finally did when we bought another IT company who had done exactly as he had wanted and suddenly he was responsible for operating such a system -- and it was exactly the disaster I had predicted (even down to the maintenance costs).   it was the only time I ever heard him apologize!

    Very simply:   You choose the tool that best gets the job done -- not the other way around -- and especially not based on some dream or ideology that is not based in reality.
    steve_jobswonkothesane
  • Reply 13 of 14
    GeorgeBMacGeorgeBMac Posts: 9,862member
    Have they tried beating the robots when they screw up?

    Just make sure the robot is not named "Hal"
  • Reply 14 of 14
    GeorgeBMacGeorgeBMac Posts: 9,862member
    dewme said:
    sunman42 said:
    lkrupp said:
    Bottom line? Don't expect iPhone/iPad/iMac/MacBook manufacturing to come to the U.S. any time soon. It will always be a pipe dream.
    As Steve Jobs explained to President Obama in 2011, it's not the labor coasts that drive where Apple manufactures its products, but the ready supply of manufacturing process engineers and access to international supply chains. I believe I recall Mr. Jobs's saying that the difference in labor costs per iPhone would be in the single digits of dollars if they were assembled in the US. We just don't have the supply of (associate's degree-level) process engineers. Makes sense, as it's only in the final assembly that any human is involved in the process.

    Frankly, this story suggests to me that (1) our robotics engineers aren't that clever and (2) industrial robotics has been focused on large system assembly (say, cars), rather than fine object assembly. New technologies will have to be researched and implemented to make the latter possible; the only question is whether it makes sense financially.

    Remember that since 2008, Mac Pros have been assembled in the US, in a largely automated plant — though fine assembly is still done by humans. I suspect the labor involved in that manufacturing process is also a negligible cost factor.... especially considering the denominator. 
    Great post, although I disagree with the middle paragraph. There’s plenty of expertise in multi axis motion control, I.e., robotics, although the US is a little behind other parts of the world. 

    The term “robotics” is way too generic to describe all of the permutations that actually involve robotics, micro level on up. 

    Size isn’t really a factor at all. Pick and place machines that populate printed circuit boards of every size have been in widespread use for decades and would qualify as robotics done on a micro scale. If you look and MEMS and Microbotics we’re talking Nano scale. 

    Automation, including robotics, is no different than any engineering discipline and practice. Some people and organizations are good at it and some aren’t. Drawing a summary conclusion about such a wide field based on any single implementation, good or bad, doesn’t make sense or paint an accurate picture of the field. 

    You can actually learn a lot about robotics programming by exploring the open source ROS (robot operating system). https://www.ros.org/


    Part of the trouble may be that our robotics engineers are mostly MBA educated in major universities.

    We lack the self-educated George Westinghouse types who are rooted in just getting the job done reality and replaced it with academics.

    In the end, a robot is nothing more than a tool to be used -- a fancy hammer.  It's not something to be worshiped.
Sign In or Register to comment.