Apple learned automation can't match human skill | Hacker News Reader

Apple learned automation can't match human skill

appleinsider.com

62 pointsKFC_Manager57 comments

I see this as more of a commentary on the lack of people able to do the automation, and the ways automation is done currently.

On a production line with 1000 different operations you'd have a thousand low skill people using their eyes, brains, and fingers to develop the assembly process.

You could never get 1000 roboticists working on getting an assembly line going. I don't know how you could even find that many, but if you did manage to do it, you'd probably end up with an assembly line that would spit out 60 iphones a second.

Instead what you might have is a setup where no individual process is getting the attention it truly needs. Each step is probably just barely working because the pressure to get 1000 automation steps working means that as soon as it works even poorly you've got to move on to the next step.

Well, that's my main opposition to "AI will replace [insert any profession]". Many jobs are so specific it will never be economical to hire a team of software+AI specialist to create and maintain the software to automate that work. Plus the fact that the said specialists usually know nothing about the domain.

There's always a way around this, and GP hinted at it. Instead of automating a difficult stage of a process as-is, you redesign the entire process, possibly including the final product, to make it automation-friendly.

This is how we deal with problems in general. If we can't solve them on their own terms, we change the terms. For example, designing an all-terrain vehicle is hard, so instead we've been beating the terrain flat, and paving it so that it stays flat, to allow a simple box with wheels to work, and we've been doing that since beginnings of recorded history.

(And speaking of paving - we're not placing random stones in the ground anymore; we're either pouring liquid, or laying stones pre-cut to standard size. Both are much easier to automate, and to some extent they already are automated.)

I can't think of a job so specific that wouldn't yield to competition from equivalent but normalized job.

A cynic could reply that you only need enough hype/marketing to convince management and a way to shift blame when shit hits the fan.

Now you need three experts, one in the domain, one for automation and an engineer. But only the engineer will be needed in production.

If lack of roboticists to spend enough time on each station was the main issue, I bet 100 roboticists would have been plenty. At that rate you can pay them a lot and still save money. And after 6 months you can reassign 90% of them to a new project, while the savings continue indefinitely.

Which is where the push to general AI comes from.

Even assuming this will be a thing one day, you still need an IT team to integrate and maintain this software.

Disappointing lack of technical depth in the report.

> Typical problems that arose include how Apple's use of glue required precision the machinery couldn't reliably match

Aren't robots commonly used to place adhesives (even replacing welds in metal fab) exactly because they can apply it far more precisely than humans? Picturing an iPad, I'd guess the issue would have been flexibility in placing glue in 3D space, around odd angles and tight corners, and not precision.

Yes, I'm sure it's related to any of the softer parts of their devices, such as being able to lay and connect the various tiny flex connectors that need a certain amount of "feel" to assemble.

And despite Apple's high volumes, with changing their product line each year I'm sure humans are much more flexible when it comes to building different devices, rather than a full robotics line that needs to be redesigned.

Yes, but it’s much more politically convenient for the incompetents in charge to blame the dumb machines than themselves for their multi-million $$$$ project collapsing on its arse.

Steve Jobs would’ve flayed the lot of them till he got exactly what he wanted. Cook’s mediocrity simply shrugs and rolls on, with nothing learned at all. #PerfectlyOiledCuckooClock

Have you ever worked in manufacturing?

Pre-press. Where in under a decade I figured out from scratch how to automate packaging artwork production successfully: by building the automation for, and putting it into the hands of, the artworkers themselves. i.e. The folks on the ground who do the actual work, and understand what that work entails.

Alas (for me), scaling that success to a $1Tn global packaging industry has proved way harder, in significant part due to the penny-pinching bean counters and talentless middle-management hacks: useless people who won’t spend a penny or commit to change when they can pay themselves the same salary for never taking a single risk or decision at all.

So what are the chances that Cook’s Apple filled a room with highly-educated experts and told them to invent a grand impressive automation solution without ever sitting with all those “little people” who make the stuff by hand in order to learn their jobs from them? From what I’ve seen of “real” applications development, and the gall of major industry vendors who rake in millions selling big-iron “solutions” that simply don’t work out on the shop floor beyond their glossy canned C-suite demos, I’d wager that’s the norm.

Ah well, now I know how Papert must’ve felt. Back t’drawing board…

I find this to be relevant to the self driving problem - Apple/Foxconn could not detect when things had gone wrong on the automation line and stop it, let alone have the line's robots fix it. However, we expect a self driving car to detect when it encounters a novel situation on the road? And it surely will.

If they could not detect it in the confines of a highly controlled factory assembly line (not manufacturing but assembly) then how can a car detect novelty on a cityscape or even a highway?

Those are two pretty different domains. In manufacturing automation, some of the things that are hard include gripping objects, handling fluids and pastes, and adapting to design changes.

For self driving cars, the controls are pretty similar, even across car models. Gas, brakes, and steering. It doesn't happen that a supplier ran out of engines and a decision is made to replace the car's propulsion by jet engines, thrust vectoring, or a hovercraft. It doesn't matter too much for self driving cars to be within 0.1 inches of the center of the road, but if your electrical components are offset by 0.1 inches or there's 0.1 fluid ounces too much glue because this batch of glue is more liquid than the previous one, the electronics probably won't work in the end.

And yet if you look at what actual experts, that aren't trying to sell anything are saying, like automotive testing bodies (EuroNCAP, ADAC, FIA, etc) the consensus is that self-driving cars are at least 15 years away. The gap between reality and what salesmen have sold to common people is gigantic.

Also the R&D budget for self-driving cars is bottomless.

There's a limit to the R&D budget for assembling THIS year's iphone, above which it's just better to use humans.

Apple manufacturing budget >> Self driving car r&d budget.

My comment was not about robotics but about detecting when things had gone wrong. Again, the manufacturing lines are in a very controlled environment - loaded with sensors, compute power, perfect lighting, humidity etc. Compare that to the noisy world of roads, traffic, weather and humans.

And in that same vein, why haven't we focused on the infinitely easier realm of rail automation and safety?

We jumped right to the hardest problem set. Probably because it's the most sensational and easiest to get broad financial support by selling people the promise of less rush hour drain.

The rail automation is solved, entire metro lines are automatic in several places. Unfortunately while it is an easier problem to solve, it's also a less financialy interesting: in a train the labor cost of the driver is much lower than in a taxi.

Replacing the cost of 1 conductor every 200 passengers is not nearly having the same impact. Moreover as others point out, it has actually been done and there are automated lines. Safety advantages are probably slim - when there's a train accident it makes news, because they're RARE.

It's also not going to significantly improve one's choice of transportation - you either have access to rail already, or you don't - regardless of who/what drives the train. Schedules are already tuned and if it was more profitable to put more trains, it would likely happen even with human conductors.

Safety regulations are what prevent more trains from running in many European places. Limiting rail to automated-only trains would allow them to run right up next to each other without the huge gaps between them that exists right now. You could move thousands more people than automated cars could manage for a fraction of the cost.

Is it really?

I thought you generally can scale trains by adding more wagons.

Only for busy point-to-point connections. You can bundle up wagons from different senders or for different recipients into one train for the bulk of the journey, but the first and last leg are often done by truck because it's more cost-effective than running trains with just one or two wagons from/to individual factories.

Now if you had a train network with autonomously driving wagons that self-assemble into convoys, that would be quite something (and probably more cost-effective than truck convoys in the long run because the lower friction of metal-on-metal compared to rubber-on-asphalt).

Yes, there's a safety limit to how close trains can be.

Sure you can make trains bigger but many times trains are already the necessary size for their time windows, destinations, etc.

Also simply making bigger trains puts more control into the hands of a few companies. If there's going to be more competition driving prices down and offering more options, then there needs to be smaller margins in safety controls to make room for more competition.

That works for goods, not so much for people.

2 examples:

1) You can't make the train longer than the station (think of a subway station)

2) I don't need 300 more seats on the 6:30 AM train, I'd like to sleep 30 minutes more and have an option at 7:00 too. The existing 7:30 train arrives too late for my needs.

to be fair, we do have some autonomous rail infra - Metro line 14 in Paris, the DLR in London, the Victoria Line was designed to be autonomous but scuppered by the unions.

The Copenhagen metro is fully driverless (no staff in any role onboard) and consequently runs 24/7.

Less sexy, less money, more government.

I wonder if there are companies focusing on solving the opposite problem. In other words, AI to focus on things humans don't perceive (leaving the driving to the human). For instance, figuring out the person in another car is drunk, then alerting you to avoid them. Or detecting emergency personnel needing you to get out of the way (can't tell you how often I see folks blocking or oblivious).

This goes along with my experience in new cars in that the best improvements are those that enhance my ability to drive (such as a backup camera).

The more I work in and watch the ML/AI space, the more I’m convinced the better approach is similar to what you describe: “augmentation” of humans’ abilities and skills rather than replacement of them.

Using advanced AI, or even a bunch of semi-decent models to condense information, highlight things humans might miss, enrich with predictions, etc so that humans don’t have to spend as much time wading through data themselves to try and extract meaning and can instead jump straight to more informed decision making seems like a better approach to me than “lol can we make a neural net that does lawyer things?”

the smaller electronic device is, the higher the precision you need. For such a compact device such as iPhone, the precision is not enough to automate the production yet. For self-driving/flying, there is always some lee-way. The demand of lowering error variance is low.

> not enough to automate the production * yet

* = economically, because low-skilled labor is toooooo cheap

Common mistake, trying to replace humans with machines. Typical of penny-pinching bean counters and talentless middle-management chair warmers. But I don’t think Apple has ever understood automation (as any 20-year AppleScript veteran can tell you), so I’m not surprised Cook’s crew has failed to capitalize.

Complex automation works best when placed in the hands of skilled humans, as an amplifier of human ability. Let them use the machines to accelerate all their mundane repetitive crap, while retaining the human ability to make reasoned decisions and handle corner cases and errors intelligently.

But perhaps a more logical place to start is by automating away the penny-pinching bean counters’ and talentless middle-management chair warmers’ jobs? Dog knows they’re reliably useless at it themselves.

> But perhaps a more logical place to start is by automating away the penny-pinching bean counters’ and talentless middle-management chair warmers’ jobs?

I readed this on HN and your comment made me remember it. https://marshallbrain.com/manna1

Heh. Yeah, I did have Brain’s speculation in mind too, but right now I think we can reasonably file its long implications under “let’s deal with that new problem when we get to it”, as right now we aren’t even close to solving the problem we’ve already got.

Personally I see successful high-level automation as much more meritocratic, in that it both serves and is directed by the same individuals: the users themselves. The worst that can happen there is that you automate yourself out of your existing job; but if you can’t think of how to parlay that win into your next then your imagination picked a funny time to fail on you now.

Assembly is mundane repetitive crap.

Elon Musk also found out something like that I guess https://techcrunch.com/2018/04/13/elon-musk-says-humans-are-...

What bothers me is that nobody realizes you can only automate simple tasks. The more complex the task, the larger the back end is to support the single device to perform the task. Sooner or later you end up with a single specific automating pieces that are costing you far more than had you just hired someone for 35k with ok benefits.

This reminds me of that "overestimate in the short term, underestimate in the long term" quote about technology.

This article is talking about the world in 2012. The world in 2012 was radically different to the world of 2020. Humans were still the best Go players on the planet, the landscape of image recognition looked rather different and the hardware was in a completely different place.

The skills to automate this stuff are developing right now. We're basically looking at a reset of these lessons that were learned in the early 2010s. The next wave has a much stronger foundation. Computers are now, potentially, better at pattern recognition. For all we know the insurmountable is currently being surmounted.

Can't fight the feeling this is just an Apple-blogger's justification for how its fine that there is a level of suffering going into the manufacture of their phone, as we know from the factories having to install suicide nets.

Bottom line is, humans are cheaper than robots in the parts of the world the make your iPhone. If you ever visit the factories where this sort of thing is made its all shockingly manual, because that's what is cheapest.

Based on what the article said, I have to ask. Is there lots of glue applied inside Apple products? Is that primarily in small devices for waterproofing? Not knowing much about manufacturing, I’d think that most people would expect parts to be secured together via solid metal screws or something other secure.

On all Apple devices the display is glued on, and on smaller devices a few of the internal parts are glued. Check out an iFixit tear down for a look at where the glue is.

Even cars are being glued nowadays. Glases, aluminum roofs, etc. There are dozens YouTube video about car body repair and glue can be found in surprising places. No solid metal screws or welding points.

Antennas, batteries, back panels, circumference of front glass panel, metal frames for the panel, etc.

> How Apple learned automation can't match human skill

Super generic title. Has it replaced all its software with humans?

That doesn't follow from the title as it stands. Automation not "matching" human skill means that automation cannot cover 100% of operations performed by humans, but it doesn't say what the percentage actually is. Could be 0% (like your rhetorical hyperbole insinuates), could be 99.999% (not yet probably), could be 80% (more likely).

It's not my rhetorical hyperbole. It's in the article title, which doesn't leave any room for the millions of tasks at which automation far exceeds human skill. That would have been "Apple learned automation can't always match human skill".

The article essentially says that if Apple can’t build reliable robots to build mobile computers more efficiently that people, then it can’t really be done. This is, of course, horsepuckey, but what do you expect from an Apple blog?

Isn't Apple the most highly valued company in the world at present? It's certainly one of the companies with the most resources to throw at the problem. That will give them a big advantage.

Throwing money doesn't solve a problem when you don't have the right people.

Took them quite a few years to fix a keyboard on their macbook line for example...

It seems to me that Apple Silicon, will eventually start with “Device on Chip” type solutions (i.e. there will no longer be a motherboard). Including vertically integrating OLED construction into the wafer as well.

Similarly, vertically integrating battery construction and housing into the device body, like TSLAs battery day announcement.

Between these integrations, I’m not sure there will be much remaining to automate.

whatever you are saying, that they will make, needs to be manufacture as well. Its not like there is iphone printer available on aliexpress.

Combining logic chips into the same wafer such that no motherboard is needed is just a matter of wafer size, owning/experience with the IP, and needing connectors for IO like speakers, mic, camera, etc. All trivial to do if you're willing to take on the costs of (or reduce) poor yield and less repair-ability. This is the current trend even for chips in AWS like Graviton.

Silicon Substrate for OLED: https://arstechnica.com/science/2020/03/blinkin-fast-oled-ma...

Details of TSLA's push to make batteries part of the device body. https://www.forbes.com/sites/jamesmorris/2020/09/22/tesla-re...

> 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.

Seriously? Did anybody even read this garbage before publishing?

Why do you find that garbage?

The human hand sensors/fine muscle control is insanely good.

I don't think we are remotely close building anything that general-purpose and that precise.

Reply on news.ycombinator.com