Places like China assemble stuff by hand that would be automated in higher wage countries not because they don't know how, or higher wage countries have developed some new software they haven't. It's because wage costs are low. As soon as it makes economical sense to automate, companies do. It's not specific to geography, or technology, or anything of other than economics of thr world labor market.
The other point that is completely missing is that a factory does not exist in splendid isolation.
If there's a problem with the FR-4 circuit boards, or the pick and place machines are misbehaving, or some of the chip-scale components that are being placed are faulty, or the wiring harness is wrong, or the molds for the body of the product need modifying because the product is cracking...
... it's much easier to deal with your supplier just down the road than one across the world.
These days stuff is manufactured in China because (the) stuff (that makes up the stuff) is manufactured in China.
Additive manufacturing is still too slow and expensive for mass production. As a recent Ars Technica article put it, it's fine for supercars but not for cars.
Yes and the solution is to have a warehouse of stuff that makes the stuff close by. But you probably need lots of manufactures to make that warehouse make sense. etc.
When I was doing my master's degree I would bike down to Rotterdam where the largest importer of piping supplies in Europe was. I just wandered off the street and every part imaginable they could just pull off the shelf for me. I was converting all kinds of wild stuff, from metric to imperial and vice versa, there are about a billion different thread standards and manufacturing standards, they had convertors for everything. And cheap stuff to go the distance.
The funny thing was, I didn't know it was there, there was an Italian researcher visiting that told me about it. However, he said that in Italy where he did his research he needed to travel over an hour to get to something similar. After that, I realized there were many pieces of my project that had large warehouses like this in Rotterdam due to the harbour. They often had tiny little windows with people with plus 20/30 years of experience, just taking orders, and coming up with solutions for your problems with a list of parts that would get you there. Then you would walk away with 50 bits, and a bill that was less than 40 euros. Madness.
I've listened to few interviews with people manufacturing in China and "you have to have someone on the ground there" repeated in pretty much all of them.
I was told by people that work for FAANG that each company reserves some number of first class seats for the daily SFO to China flights specifically for this reason.
If there an issue, some engineer responsible for the part can be in the factory in under 24h
Meta (Quest 2/3/Pro, Meta Portal), Google (Pixel, home automation hardware, much more), Amazon (Echo, Kindle, Ring), and Microsoft (Surface lineup and much more) all manufacture hardware.
The "no, _they're_ the idiots" rivalry between engineer / manufacturing is almost as old as time. A design team almost never gets to just "hand off" the files to the manufacturing side without some amount of issue/fix iteration happening.
It's not just the physical engineering of a device, though. If the $thing that is responsible for QA/Test or even programming keys into $widget goes down, the factory has to stop and that's _expensive_ so you get the guy that designed it out there to fix it _asap_.
>Places like China assemble stuff by hand that would be automated in higher wage countries not because they don't know how, or higher wage countries have developed some new software they haven't. It's because wage costs are low.
Not only. It's also because important parts can't be automated with current technology.
It's also not that Chinese factories substitute robots for manual workers. They automate the shit out of them too, and have the same share of robots vs workers that a western factory would...
Without further context, we cannot say whether a lower number of robots in the US is an indication of inferiority or superiority to China.
Example: A company I was working for produced booklets with their own autmatic book printing binding machine. Not long ago, they outsourced it to a printer because it was cheaper that way. So the single more capable printer's "robot" is replacing the need for a lot of other not so efficient small "robots".
It's not China manufacturing iPhone, its Apple and its partners.
So your first sentence should not have China as a subject.
>Without further context, we cannot say whether a lower number of robots in the US is an indication of inferiority or superiority to China.
Its not a national defence thing. It's not the US vs China. It's Apple. Apple US, Apple China, same thing, still Apple.
Apple makes a factory, that factory has certain economics based on its inputs (labor, materials) and outputs (iphones). Nowhere does China come into play.
We live in a globalized world, it is run by large rich companies, not governments.
I don't see that I am missing the point. My point was that without further context the numbers can be interpreted either way. You were at least trying to providing a context. So I see no contradiction here. (NB: "Chinas" was the subject of the parent's remark.)
Agree, in that if the cheapest way to do process is automated and the capital cost of the automation is reasonable there is no advantage to offshoring that process.
An example is pick and place assembly of circuit boards. For large quantities, meaning the loading of the feeders is not a big part of the cost, a local company can match or undercut the price of an offshore company.
A counterexample is chip fabrication, where the capital cost dominates (and even if it didn't wafers are often manually handled).
The percentage unit cost and environmental impact of shipping electronics from China (or between any two points on earth near a major harbor) on a cargo vessel is zero.
If the labor in China is cheaper than the robot + the person that operates the robot, then it is cheaper to build it there.
Building it locally only makes economic sense if the design is changing frequently (e.g., using local pick and place for prototyping), or if it is built to order, so the time it spends on the boat actually matters.
As an extreme example, giant Italian windows cost about the same, delivered, in the US as inferior domestic ones do. The main downside is the lead time. In particular, thanks to a drought, traffic through the panama canal is all screwed up these days, so getting them to the west coast is hard.
If the economics tilt this way for shipping giant panes of glass, they certainly also do for tiny, high priced electronic devices.
Of course, externalities (e.g., environmental impact of Chinese manufacturing, human rights, national security and xenophobia) could change the tradeoffs.
Modern container cargo ships are huge. Even though in aggregate the absolute pollution is a big numbers, when you amortize over the volume, shipping an item has about a very close to zero impact on environment.
And this is not some hand-waving explanation either, since other factors dominates the environmental impact (for example, shipping from Newegg to your office is probably a magnitude more polluted), it doesn't matter if the factory is in China or the US. This applies to a lot more things, even lower value stuffs like fruit. Which is why local produce might have more environmental impact than imported mass produce fruits, though I don't have a citation for this on hand.
"According to Irene Blooming, a spokeswoman for the European environmental coalition Seas at Risk, the fuel used in oil tankers and container ships is high in sulfur and cheaper to buy compared to the fuel used for domestic land use. "A ship lets out around 50 times more sulfur than a lorry per tonne of cargo carried."[1]
"Of total global air emissions, marine shipping accounts for 18 to 30 percent of the nitrogen oxides and 9% of the sulfur oxides."[1]
"One source of environmental stresses on maritime vessels recently has come from states and localities, as they assess the contribution of commercial marine vessels to regional air quality problems when ships are docked at port.[33] For instance, large marine diesel engines are believed to contribute 7 percent of mobile source nitrogen oxide emissions in Baton Rouge and New Orleans, Louisiana. Ships can also have a significant impact in areas without large commercial ports: they contribute about 37 percent of total area nitrogen oxide emissions in the Santa Barbara, California area, and that percentage is expected to increase to 61 percent by 2015."[1]
"
Maritime transport accounts for 3.5% to 4% of all greenhouse gas emissions, primarily carbon dioxide.[1][27] According to the World Bank, in 2022, the shipping industry's 3% of global greenhouse gas emissions make it "the sixth largest greenhouse gas emitter worldwide, ranking between Japan and Germany."[1]
While this is true, if you were to calculate the last-mile carbon cost it would be much, much higher.
A full container of goods costs about 60 gallons of diesel to ship across the world.
That same container costs 60 gallons to move 360 miles from a port.
Delivery vehicles drive approximately 0.8 miles per package delivered, and get around 10-20mpg delivering, so each package costs 1/10 to 1/20 of a gallon of fuel to deliver from the depot to your door. Assuming there are 4000 eventual packages in a container, the last mile delivery for the container is 200-400 gallons.
Obviously this is not the case where the final destination is a large receiver, but even then there are still significant transportation impacts in most cases.
I hate the fact that when I read the part about sulfur dioxide I’m like, well, that’s good at least lol. There’s some evidence that cleaning up ship emissions might be a significant contributor to recent sea temperature acceleration, since the cloud seeding effect from ocean shipping has been drastically reduced.
End of page 129, start of page 130 (page 3 and 4 in pdf file).
Truck used twice as much energy consumption as all marine vessels. Dry cargo (there are a bunch of ships shipping oils and LNG too) is generally around 60% of marine vessels.
They hurt recently phased out the sulfur fuel, and it's a whole ironic thing, it's actually causing Temps to go up because the sulfur clouds were actually helping us.
Automation sits in a lot of different positions: sometimes it's a cheaper option overall, sometimes it's actually more expensive but more flexible than manual assembly, and it varies greatly from operation to operation (some almost exclusively manual and some almost exclusively automated)
There is such a thing as pressure to automate. You could imagine a world where wage costs just stay low, and companies just keep making stuff by hand. Companies need to know that there is a cost-benefit trade to be made.
And so an article like this is nodding at that pressure from the opposite direction:
"Hey, you should be aware that automation (cost savings) are totally feasible, even if they're not realistic for you at this exact moment."
And the article only mentions reduction of costs with automation. Would be strange enough for some tech-oriented magazine, but this one is called "The Economist". Though maybe it is supposed to be read just as a futuristic speculation, or as an overview for people unaware that automation is possible.
Yeah. China is full of make-work programs. It's obvious.
For example, every Chinese city has sidewalks made, not of poured concrete, but rather 30 cm x 30 cm ceramic tiles. Due to historic city planning, the Chinese tend to park their cars on the sidewalks, which inevitably breaks the tiles, thus causing regular replacement. These tiles are laid not by some sort of brick laying machine, but rather by hand.
The sidewalk by my house was poured in 1968. I know that, because there's a date stamped in it. A couple of blocks over there's a sidewalk that someone scrawled graffiti into. "Elton John 75". There's literally nothing wrong with either of these sidewalks. They're not cracked, nor buckled.
China doesn't need to replace every sidewalk every two or three years, but they choose to.
Truthfully, based on the state of our sidewalks in the richest cities of the US like NYC, SF, Chicago, and LA— we need to be replacing them maybe not every two or three years, but at least every 5. Right now it seems like 20, if we’re lucky. Meanwhile most towns and newer parts of smaller, less affluent cities, or even wealthier suburbs don’t even _have_ sidewalks.
I think it’s ultimately part of our philosophy toward the built environment that “Europe is built with screws, the US is built with nails”— we build something that gets the job done at the time it’s built, then never come back to it. Maybe that allowed us to build fast, but now the bigger issue is maintenance of what we already have.
nails are made of a more pliable metal, whereas screws are more brittle.
the real "long-term" fastener, in my opinion, would be bolts; but I don't see anyone using those to frame a house.
just a counter-point, if it is at all; I grew up in SoCal with poured sidewalks around the entire area. Nearly every single one of them was destroyed by nearby trees or shrubbery extending their roots under the thing and cracked it up into a million pedestrian-tripping pieces.
These sidewalks were from the 60s and 70s too -- thing is that they cracked 20-30 years ago and no one ever bothered to replace. The home owners i've spoken to that wish they could were confronted with the aspect of being forced to pay for their entire street lest they wish to pull permits to repair their small section.
So you have this new problem with a liability tug-of-war between well-wishing civillians against the cost-saving municipalities, and to be clear this was more than one single municipality, i've seen it across SoCal.
So, what i'm trying to get across is that there is a middle ground between 'needless replacement for the sake of generating work' and 'neglect', and neither China or the US has seemed to pin that middle down yet : but that needless replacement regime sure sounds nice as a pedestrian that just wants to use the goddamn sidewalk. I've seen busy-work before, that just doesn't strike me as that.
Here in the netherlands we have a tremendous amount of sidewalks per capita and they are 99.99% tiled. Some of the sidewalks are over 50 years old and stil in good shape.
There are many benefits of tiled sidewalks. Its pretty easy to put cables underneath. Happens every few years (internet, electricity). You also dont need fancy equipment to make them. Or repair them.
Dutch tiled sidewalks are different from what I have seen in China. They embed thick tiles, hardened bricks nowadays, in sand. Chinese make concrete sidewalks topped with thinner tiles. Nice when it rains some tiles work like mud pumps spraying your feet.
BTW I have no statistics on Chinese sidewalk maintenance, but when I lived there development was so fast that a sidewalk would have to be broken up every 2, 3 years anyway for new cables, pipes or whatever. I always wondered why they didn't use the tile plus sand system, as this involved always in noisy and dusty concrete hammering.
A contributing factor is also that they are a Communist State with a lot of people that need jobs!
You could argue that Full Employment is tenet of Communism.
I'll leave debate about what Communism is and what the CCCP is and isn't to the rest of the internet. I'm sure there are people out there who would love to have a conversation about it... LOL ;-)
One of the best things about doing business and also non monetary interactions globally is opting out of local distractions
As soon as you unsubscribe to US perspectives, nobody is detailing conversations about the PRC ultimately winning that territory and the ways it behaves, it is very easy to operate there on a day to day basis
They’re a state that in theory is attempting to move towards communism - china’s government is not communist yet. Actual communism would be a much different looking country.
Also, communism and automation _absolutely_ go hand in hand. Both socialism and communism have no absolute need for employment, that is a very capitalist view. The requirement for communism and socialism is that every one’s needs are met. So long as the state can provide food, shelter, etc, the only reason people should have to work is to help meet those goals through production.
It’s certainly true that the state has different behaviour when ruled by the working class (like in China, Vietnam, Cuba) vs the capitalist class (like in most other countries).
That doesn’t necessarily mean everything is centrally planned vs everything is a market. A variety of approaches can be used to satisfy the interests of the ruling class, within the constraints they have.
So in China there is definitely state intervention to ensure there are enough good jobs. But that can also mean speeding up development and automating so the jobs are better, while creating new jobs by producing new things society needs or wants.
There was a revolution led by the working class. The material conditions of the working class have been improving since, while in similarly developed capitalist countries the reverse is true. Banks and land are all collectively nationally owned. Capital controls are strong enough that not even the richest can move too much of their wealth outside the country. Capitalists are subject to the law frequently, even billionaires are imprisoned.
I’m not a chinaphobe at all but that’s a rich interpretation
Communism relies on 2 phases, the first phase requires all power and resources consolidation in a standing committee, the second phase then allows for the collective communist nirvana. The second has never happened anywhere.
The standing committee realized that and switched away from attempting that, veering to state capitalism to bring a controlled set of benefits from capitalist drive. They still teach that they are doing marxism but there is nothing working class about the party.
One way to corroborate this is to look at what the Tianemen Square protesters wanted. They wanted democracy: to democratically vote in more communism. The standing committee is not interested in that.
People that fall for “true communism” - phase 2 - always get disillusioned when they notice the wool pulled over their eyes by a select few taking all resources for themselves, instead of for “collective” ownership by the workers.
Or alternatively, the events in 89 were part of a wider trend in socialist countries of NATO-backed violent uprisings with merely rhetoric of democracy.
For example, I have spoken to many Romanians that lived through the 89 events who see it as a coup against socialist democracy. It succeeded in România, it failed in China.
Your brain has been melted by US media I'm afraid. Using 20th century soviet style communism as your mental model for modern China won't serve you well
> “Thirty years from now we will laugh at our generation of humans, putting products together by hand,”
I never laughed at humans thirty years ago even if they were programming on antiquated machines with only megabytes of RAM over dialup internet connections. As a matter of fact, I think some of them are legends and we should try to emulate them.
I don't know why humans thirty years from now will laugh at us, even if it happens that robots do more of the jobs we do take for granted.
It's even more likely that, if automation improves as they foresee, 30 years from now we would not be laughing, but crying, while trying to make do with meager welfare and living in slums, after we've been made redundant.
I'd argue it wouldn't be profitable for world elites to manufacture products for the masses that they couldn't afford... but if we live in a future where manufacturing anything is an easy task, and software can dream up realistic fantasies, then why would the elites even need the masses at all.
Right and assembling things by hand will probably become a niche hobby like retro computing. For small scale startups assembling things by hand may still even be a valuable skill in some cases.
Thoygh with the way AI is progressing we may still have cheap humans doing tedious tasks while the AIs make art, program, and manage companies for the benefit of AI wallstreet and digital AI avatars of dead billionares. :p
As an aside, I don't know why the Americans make the DMV sound like it's so much bother. I've read so many complains about this.
In other countries I know periodic car inspection is like a few hours at most, every couple of years or so (and that only starts after the car is a good few years old already).
American here. The DMV (sometimes the BMV) is a state-based agency, where state is a regional authority, as opposed to federal, which is a national authority. California’s rules and requirements on when you need to visit the DMV are different than Ohio’s rules for its BMV.
Still, you generally never visit for vehicle inspections. Third parties take care of that.
You visit the state agency office for licensing issues. Want to drive a vehicle? Want your vehicle to be properly registered in your state? Go to the DMV.
For these essential functions, you should make an appointment. But truthfully, the appointment doesn’t really matter. There are too few offices for the demand, so most space in the agency is dedicated to the waiting room. I’ve waited nearly a day one unlucky time. And there are different functions per line. Your form may need to be processed by multiple functions. So you need to wait in multiple lines. It is a few minutes if no one is there, but hours otherwise.
The only reason why you’re there is because you have a relatively simple thing to accomplish. And you’ll only need to do this a few times while you live in that state, since renewals are usually done via post.
Is it the worst thing we have to do? Nope. But it’s highly disruptive because you don’t know how long it will take, despite being relatively straightforward. It’s an example of poor customer service’s power to shape opinion.
The identification requirements are strenuous in my state. Perhaps to combat fraud and out-of-state registrations, e.g. to get cheaper insurance. I needed my birth certificate, a utility bill from within 30 days, and my passport last time I renewed my license!
It’s one of those rare government offices that pretty much every American adult has to deal with at one time or another. I agree that it really isn’t a big deal.
In the past decade, I’ve had to visit the DMV office twice for a combined total of about two and a half hours. I’ve also had to get my two cars emissions tested about three times taking about 30 minutes each time. 4 hours of my time over a decade isn’t worth getting worked up about. It’s annoying but so are about 1,000 other things in life.
For normal cars here (Italy) inspection is first time after 4 years and then every 2 years.
Once (but frequency of inspections was lower if I recall correctly) you could only have it inspected by the equivalent of the DMV, since several years there are authorized (private) workshop/mechanics that can do it, so the process is much easier/faster than before and you have the possibility to choose among many different places to go and it is usually a matter of a few days at most to get an appointment.
For some kind of papers/authorizations (as an example to add a trailer to your car or if you import a car from another country) you still need to actually go to the government offices, and here - usually - the troubles start, appointments only available after several months, long queues anyway, even minor issues with the documentation resulting in stopping the procedure and needing a new appointment, a nightmare.
Recently a friend bought a new car and needed to have the modifications for his disabilities (made by a specialized, authorized workshop) inspected and approved, it took him (actually to the workshop, that AFAICU has even some advantages in the queuing system) almost four months to get the vehicle inspected and approved.
These articles are written by people who have no idea what they are talking about and have never built anything, much less anything in the physical world before.
The article doesn't distinguish between making parts and assembly. Making parts was mostly mechanized decades ago. The article talks about 3D printing in metal, which is a nice technology but way too slow to be used to bang out consumer products.
Assembly is still tough to automate. If you want automated assembly, you usually have to design things so that the parts go into place with little or no fuss. Vertical assembly is common, where everything is placed with a straight-down move. This goes way back, to the first cheap mass-produced clocks, where there's a baseplate with pins sticking out of it, you just drop on the gears in the right order, and it all comes together.
They mention Bright Machines. Their site is heavy on the hype, weak on the machine specs.[1]
They sell some kind of robotic assembly cell. So do many other companies. Sony had a nice one thirty years ago for making the Walkman. It's not clear what Bright brings to the party. More software integration, probably; they talk about that a lot.
There are some new results from MIT on assembly planning.[2] This is kind of cute. They start with a game-type physics model of the assembled item and figure out how to take it apart. That's reversed into an assembly plan. Joint work from MIT, Texas A&M, and Autodesk.
It's surprising how much manual work is involved in assembling a mobile phone. It's a high-volume product with a modest number of parts, the ideal case for automation. Many early US-made mobile phones were assembled automatically. They were stacks of boards that just had to be dropped onto alignment pins and pressed together. Newer phones have tiny flat cables and connectors all over the place.
Most automated assembly lines involve a huge amount of custom tooling. It's not a few robots with standard tools.
The state of the art in robotic assembly is far worse that you'd think it would be. And that's after half a century of work on the problem, all the way back to McCarthy in the 1960s.
Most people don’t understand how fast tolerances can stack up during assembly and how complicated any step could be. Oftentimes for complex parts you may need to articulate on 5-6 axis or support in a complex orientation or just shim out existing mismatch. Humans are really cheap compared to robots that could do the same thing.
But, they will mainly be read by people who have no idea what they are talking about and will never build anything, much less anything in the physical world...so, it all balances out I guess?
Can confirm both. I have a composites design & fabrication biz, and while I always seek good automated solutions, and indeed some parts can be automated, we are soooo far away from what that article propounds that it is a ridiculous fiction for any serious product development & mfg. Maybe might work for some cheap Chinese-junk type stuff, but nothing real, at least soon.
The problem is that some of the ignorant writers and readers don't recognize their own ignorance, and are in a position to make corp & govt policy based on that ignorance.
Much of this discussion is swallowed up in two differing definitions of "design." They contrast chip design with leafblower design. Chips are very fussy and require lots of patience and intellect to be useful to people, while the latter's success is more rooted in the aesthetics, ergonomics, user needs of everyday families. This isn't saying they're wrong, just a sloppy way to talk about it.
What they're really saying is the path from CAD to CAM is shorter than ever, and more integrated. It sounds pretty great. It does make me curious about where the tooling cost outstretches the gap between family and factory owner. Probably making lots of Temu-grade consumer goods, such as a disposable lawn mower, or a coffee maker that's good for 6 months. Which is of course horrific for long-term ecology. It would be nice to see this amount of cleverness directed at extracting usable materials from landfills :) :) :)
Usually technology bumps out the lowest class workers, but genAI is taking a bite right out of the middle to upper class... certain executives, office salespeople, some programmers, actors, and such will be replaced by robots before garbage workers, lawn services or roofers.
"It is difficult to get a man to understand something, when his salary depends upon his not understanding it." -- Upton Sinclair
It's my experience that coders, software engineers, accountants, and other STEM professionals regularly dismiss that their jobs are threatened by the encroachment of advances in AI such as deep learning and LLMs. And, unfortunately, the standard optimistic, hand-waving refrain of "There will be other jobs." denies reality that they may well make less money elsewhere, be under/unemployed, or not be able to retrain. I foresee a widening excess supply of skilled individuals who will face difficulties finding any job that either: pays enough OR where a potential employer doesn't deem them "overqualified." (In addition, almost all of the professional class bets against their own interests by being predominantly anti-union, even when they're under/haphazardly paid or overworked.)
It will take more decades, but even nursing and care will be automatable when AI and mechatronics advance sufficiently. I look forward to the kitchen and bathroom cleaning robots. These are moonshot projects that require 100M-1B and multiple decades to attain.
I am optimistic for all those professions you listed, because those are all knowledge workers.
I like that term because the vagueness is actually helpful. Ultimately these people work to become experts in various knowledge systems.
And the other thing that’s happening while AI is advancing, is that the educational system is borderline collapsing.
So what that means is that, even if AI automates a ton more coding, even if it’s capabilities keep improving, the net effect of that will still be to add more complexity: and that means the jobs of people who are doing that type of job now, are going to get more complex.
Someone has to make sure the AI data lake processor is feeding data into the chat room that displays daily metrics, for example. Now take that example and multiply it by 100, for all the new capabilities AI will create.
So maybe you don’t write the app necessarily, because AI does, but because of the fragmented nature of various vendors and services, someone still has to make sure the app talks to the DB, which talks to the AI chat bot, which does a new kind of real time metrics analysis made possible by AI. I really don’t think the demand for the type of well educated worker it takes to manage all that is being met, and won’t be for a long time to come. If you can do that work, your job will be safe.
Unfortunately, knowledge workers aren't a uniform hivemind, so your presumption is overly broad. Firstly, there is power law variance in capability for higher-ordered reasoning.
As automation increases and self-modifying systems are deployed, there is a risk of losing grasps on simplicity and understandable systems. There are increasing possibilities of clever obsfucation by complexity for intentional purposes of job security and vendor lock-in and unintentionally for facilitating machine (self-)manipulation (that sounds dirty but you know what I mean).
The number of high-income jobs is going to contract, but perhaps the high end will increase in pay in very limited numbers. I'm sure that the high end of IC7-9 will increase, perhaps with pay exceeding $10M USD/year in a few cases. I already know of people making $800-1.2M.
And you still need people to maintain and repair the robots, at least for now. Software doesn't need maintenance in that sense; it doesn't have moving parts that need lubrication, adjustments, and that eventually wear out.
As anyone who’s ever maintained software can tell you, software does suffer bitrot over time. Dependencies need updating and replacing. Deprecated functions have to be replaced. Systems and software change over time, breaking your software.
It’s not as high maintenance as most physical things with moving parts though.
Agreed. It’s really astounding how much maintenance “modern” software needs. A lot of the maintenance is to do with the dependencies, and/or with the discovery of vulnerabilities, but also with obsolescence in the tool chain used for building the software, or the APIs that the software depends upon, or simply the continual replacement of tokens/certificate and so on that have built in expiry.
It’s not glamorous work, but it is absolutely relentless, endless, tedious stuff.
Also it doesn't murder itself and what is around it in case of bad inputs or failure. The failure modes of software are pretty tame, specially in cloud.
With proper sized robots killing someone or damaging itself and other things is entirely real possibility.
On embedded system software can murder itself relatively easily.
* Wrong EEPROM emulation routine is using data flash (100k of writes) as EEPROM (10M of writes).
* If you make a mistake during bootloader setup, bootloader can delete whole flash memory, including bootloader, or write over the bootloader.
* You can unpack program into RAM (i.e. Compressed file on flash -> Unpack into RAM), so if you will screw up pointers, you can overwrite your own program on the fly without even noticing it.
There is a more academical definition: a robot is a computer with sensors and actuators attached. Self driving cars certanly fit that criteria.
There is a more practical definition too: a robot is a machine with a computer which does not work yet. Because once it starts working people stop calling it a robot.
For example a modern washing machine certainly fits the academic definition of being a robot, but most people wouldn’t call it one.
In hindsight it seems that way, but until recently, getting a robot to generate a movie or act as a creative therapist seemed impossible, not a matter of engineering resources.
Yeap. The VA commissioned Dean Kamen to build the Luke Arm, a practical prosthetic arm for disabled vets. It took years, million$, determination, and extensive testing to achieve. Not impossible, but it's not something that can be thrown together in a weekend hackathon.
That's fascinating. I'd not thought of it that way, but it is absolutely true. When DALLE2 came along last year I was stunned as it seemed like we had skipped the part where AI did the drudgery and jumped right to the bit where it emulated human emotion and imagination.
I know many people using AI to write vast swathes of content, to create marketing materials, to write code and to create stock images. Some of the people AI has replaced were taking $150K salaries and now are earning zero. That's a harder lump to swallow than your working class person who might have gone from $20K to zero.
Give robot a box of Ikea furniture. Even when it is designed by computer and most of the pieces are built by CNC machines (technically robots) I don't think that there is any robot in the world, which will be able to build the furniture together.
i read it more as a "horses -> cars" thing, like people still riding on horses will be soon eclipsed by people in cars, for all the efficiency reasons cars ultimately won in the end
place your bets? i went horseback riding once and had a great time...
Places like China assemble stuff by hand that would be automated in higher wage countries not because they don't know how, or higher wage countries have developed some new software they haven't. It's because wage costs are low. As soon as it makes economical sense to automate, companies do. It's not specific to geography, or technology, or anything of other than economics of thr world labor market.