Human labor is the new buggy-whip

March 14, 2025 9:37 AM

America Is Missing The New Labor Economy – China's Dominance Playbook, General Purpose Robotics Is The Holy Grail, Robotic Systems Breakdown, Supply Chain Hardships, The West
posted by Fupped Duck (27 comments total)

Comparing a robotic system to a human, the current labor force is lower skilled, lower ability, and a much higher attrition rate.

No.
posted by Headfullofair at 9:46 AM on March 14 [3 favorites]

The implications for warfare are especially chilling...
posted by Fupped Duck at 9:51 AM on March 14 [2 favorites]

We are in the early precipice of a nonlinear transformation in industrial society, but the bedrock the US is standing on is shaky.

I'm not the most careful reader, but the repetition of this sentence in the article kind of jumps out at me and makes me suspicious
posted by ginger.beef at 9:55 AM on March 14 [3 favorites]

All I can think is thank God it's a socialist country leading the way on this. Luckily the West is too busy imploding to notice.
posted by Space Coyote at 10:10 AM on March 14 [1 favorite]

As the AI bubble begins to collapse, its cheerleaders resort to the old standby of fear.
posted by NoxAeternum at 10:16 AM on March 14 [4 favorites]

Chinese robots are good and getting better. But Chinese firms catching up to existing automation equipment from OECD nations is not a new era in manufacturing. As ginger.beef points out, this article just keeps repeating the claim of a new era dawning without providing any supporting evidence.

While indigenous Chinese automation will certainly displace imported automation systems, it isn’t clear that Chinese firms will develop the sort of support networks necessary to displace, for instance, PLC controllers on the global market. Or that Chinese systems are offering some new level of capability.

Show me a robot that can sew a t-shirt and I’ll change my mind.
posted by Headfullofair at 10:16 AM on March 14 [8 favorites]

My test use case is turning a hotel room between guests. Easy, or hard?
posted by torokunai2 at 10:19 AM on March 14 [3 favorites]

Robotics are 100% going to supersede many categories of unskilled and semiskilled human labor.

However, the notion that this will afford unprecedented comparative advantage to China is very dubious.

Chinese manufacturing currently prospers at the sweet spot of developing market costs (especially for labor) and first-world professional, financial, logistics and infrastructure resources.

However, there is no reason to suppose that robots will be cheaper to develop and operate in China than elsewhere or that China will have better robots than those which can be put into service elsewhere.

Indeed, there is good reason to believe robotics will not only reduce China's direct manufacturing cost advantages, but will be broadly deployed in services in the US and Europe and lower indirect costs for manufacturing there too.
posted by MattD at 10:31 AM on March 14 [3 favorites]

The article was all about industrial robots, so i dont know why people are talking about hotel robots or shirt making robots?

China's largest state-owned shipbuilder, the China State Shipbuilding Corporation (CSSC), built more commercial vessels by tonnage in 2024 than the entire U.S. shipbuilding industry had since WW2, the CSIS said.

What western corporation would use their capital to invest in this compared to more stock buybacks?
Germany? They are already shutting down many of their heavy industries because of high energy costs. Robots won't help with that.
posted by Iax at 10:50 AM on March 14 [3 favorites]

The site is called semi-analysis, and that’s exactly what you get. It seems to have been half written by an inattentive AI, despite the team of authors. Useful only in the sense that you can see what brainwormed web 3 guys are thinking about now, otherwise pretty weak
posted by The River Ivel at 11:08 AM on March 14 [4 favorites]

People who have worked so-called "unskilled labor" jobs or even read much about them know that all labor is skilled labor.
posted by SaltySalticid at 11:23 AM on March 14 [17 favorites]

Shirt-making robots are industrial robots, and it's actually a great example of the key near-term problems facing production robotics in perception (the way a seamster/seamstress uses both vision and touch to guide cutting and sewing, even at industrial scale) and dexterity (the way a seamster/seamstress uses their hands to manage the work quickly and at low error rates).

Even with Chinese labor costs, if you could do a 1:1 substitution of a robot for 8,000 hours a year of human seamster/seamstress, you'd profit even that robot cost you $40,000 a year in capex and opex ... but no one has yet made a robot capable of that substitution.
posted by MattD at 11:46 AM on March 14 [5 favorites]

of human seamster/seamstress, you'd profit even that robot cost you $40,000 a year in capex and opex ... but no one has yet made a robot capable of that substitution.

Per memory of a Peter Zeihan youtube video there is a t-shirt automation solution which 'makes the shirt' for under $1. Memory said $0.25 a shirt and looking up that 2 months ago a PR statement was $1.05.

I'm not sure what your criteria is to make your statement but there does seem to be an automated clothes making solution.
posted by rough ashlar at 11:55 AM on March 14 [2 favorites]

This opens by shitting in my eyes with an AI-spawned image, and is on a site with a pop-up urging me to subscribe to their AI news, so I assume all the body copy is gonna be AI blather and close the tab immediately.

I don't think "roboticize everything" is a "labor economy", if anything that's an anti- labor economy. "Roboticize everything" is a lack of an economy as long as all the profits go to the robot owners, who will shortly discover that nobody has money to buy their crap.

Nationalize every AI company, distribute their profits to everyone equally. All their shit was based on ignoring copyright so none of it should belong to them.
posted by egypturnash at 12:00 PM on March 14 [10 favorites]

AI is the miracle tool that when youre eating breakfast it can tell you that you are eating breakfast, usually correctly*

* stolen from a blue sky commentator, and I just like it, so don't spam me AI boosters. 95% of us think it sucks and find it highly suspicious the atrocious FINTECH goliaths are shoving it down our throats.
posted by WatTylerJr at 12:05 PM on March 14 [3 favorites]

> rough ashlar: "Per memory of a Peter Zeihan youtube video there is a t-shirt automation solution which 'makes the shirt' for under $1. Memory said $0.25 a shirt and looking up that 2 months ago a PR statement was $1.05.

Do you have a link to this? I wasn't able to find what this was referencing from a cursory skim of Zeihan's YT page. If you don't have the video itself handy, do you happen to remember what this t-shirt automation solution looked like? Last time I looked into textile automation many years ago, they seemed pretty far away from something that could sew shirts at scale and I'm curious about how they managed to figure it out.
posted by mhum at 12:16 PM on March 14 [1 favorite]

There are plenty of aspects of garment manufacturing that are highly automated, including weaving, knitting and cutting yardage, but pulling work around and through sewing machines isn't one of them yet ... robots don't have the eye, feel or dexterity to handle fabric yet.
posted by MattD at 12:39 PM on March 14 [1 favorite]

LMAO. We can't even keep the Shake Machines at McDonalds running. Can't wait for the hilarity of Internet of Things that Look Like People.

Apropos of nothing; yesterday I was driving down a major thoroughfare during rush hour in LA, when a Waymo decided to turn left across oncoming traffic in the middle of the block (to nowhere, there was nothing to turn to). People (people!) managed to swerve out of it's way and avoid collision, but the thing was now wedged across one and a half lanes of traffic, and it's rush hour, it's not going anywhere.

Posted a video of it wedged in a discord, and my friend said he had one blow through a stop sign and cut him off. I saw another one (same thoroughfare) that froze when it encountered a flashing red traffic light (blocking off an entire lane).
posted by chromecow at 12:41 PM on March 14 [4 favorites]

cursory skim of Zeihan's YT page.

Well that wasn't gonna work because he made the comments a while ago. Perhaps over a year even and the title wasn't "$0.25 automated tshirts" but a 1-2 sentence reference about his automation worries/not worried.

There used to be https://ytks.app/ but one would have to pick a couple of tools and roll your own. The youtube DLer to glob the whole channel to get a list of URLs in the 'already DLed' recordkeeping. A bit of massage and that feeds into a javascript that if you have the URL it'll DL the transcript. From there you grep the DLed transcripts.

Do you have a link to this?

i've kinda stopped bothering with providing links because there was no benefit 'round these parts. But you did ask.

*snicker*

AI Overview
Learn more
T-shirt manufacturing automation, while requiring significant upfront investment, can significantly reduce labor costs, with some fully automated lines achieving costs as low as 33 cents per shirt,nd and potentially enabling production in regions with higher labor costs.

one of the source links for the above "AI" response. That link is also from 2017 and mentions pants as automated.

A company called SoftWare seems to be a leader in automated garment so you can dig a bit more there.

( AI company) ... All their shit was based on ignoring copyright so none of it should belong to them.

Except "all" isn't true. The image processing for radiology isn't such 'ignore copyright' position. Under this 'ban it all' position - ya axing AI for radiology?

Are you using audio transcription on your computer? If so that's what we are lumping under AI - so if you are blanket hating AI consider checking yourself before ya wreck yourself.


LMAO. We can't even keep the Shake Machines at McDonalds running.

Isn't there a lawsuit over the captive marketplace repair of those machines represents? The failure rate of those machines may very well be a revenue generation reason. But by all means lets blanket trash AI because a revenue model based on breakage is sure the fault of AI and not at all the C-suite humans.

Like how Ford now has plastic oil pans which are 1 time use. Anti-consumer can be anti-consumer without it being 'because AI'.
posted by rough ashlar at 1:17 PM on March 14 [1 favorite]

Except "all" isn't true. The image processing for radiology isn't such 'ignore copyright' position. Under this 'ban it all' position - ya axing AI for radiology?

Thinking nationalization and profit distribution is a ban is your first problem
posted by We put our faith in Blast Hardcheese at 2:09 PM on March 14 [4 favorites]

The image processing for radiology isn't such 'ignore copyright' position.

I can’t speak for any particular bespoke convolution-based medical software package but I have terrible news for you about OpenAI’s CLIP and its many, many descendants (including basically all diffusion models used for image generation).

Nationalize every AI company, distribute their profits to everyone equally. All their shit was based on ignoring copyright so none of it should belong to them.

Fully agreed. DeepSeek’s “here are our model weights and the software stack we use to train and run LLMs ten times more efficiently, fully open sourced” is a start but is not sufficient. If you use any copyrighted material you should be required to provide a full and complete index of all material used during the entire training process, and make available any material not accessible to the general public for verification and duplication.

“But Proprietary data is in our trai-” then RAG that shit and move on. If you absolutely have to use legitimately proprietary shit in your fine-tuning, generate a bunch of synthetic data based on the real-world formatting / templates / schema and include it in your public-facing training set.
posted by Ryvar at 2:16 PM on March 14 [2 favorites]

I spent a minute looking and found SoftWear Automation for those interested. The site talks about automated garment factories, but I haven't been through it in any detail.
posted by rhamphorhynchus at 2:38 PM on March 14

> rough ashlar: "A company called SoftWare seems to be a leader in automated garment so you can dig a bit more there."

This appears to be the company in question, SoftWear Automation. Unfortunately, it would seem that the claim that there exists a t-shirt automation solution may be slightly over-stated. According to this 2021 Wired article (archive.is link),

The company launched in 2012 with help from the Georgia Tech Advanced Technology Development Center and a contract with Darpa. Two years later, a prototype was up and running. By 2017 work began on developing a production line that could mass-produce shirts. That same year, the company struck a deal with a Chinese apparel manufacturer to set up a large production facility in Arkansas. That deal fell through, though, and SoftWear is now focused on opening its own garment factories.

[...]

Since its initial funding, SoftWear has raised $30 million in venture investments and grants—including a $2 million grant from the Walmart Foundation. Rajan says it will take tens of millions more to get production to 1 billion T-shirts per year. To reach that target, the company will need multiple facilities, each with its own Sewbots and skilled workers to maintain them. Rajan says a Sewbot work line can make a T-shirt every 50 seconds. At that rate, if run continuously, one work line could produce just over 620,000 T-shirts per year—meaning it would take 1,607 Sewbots working continuously to reach 1 billion in a year. Rajan says a more realistic number is closer to 2,000; so far, the company has made fewer than 50.

It appears what we have here is, at best, a potentially promising approach from a startup that has yet to make significant inroads into actual production. I would note that the most recent news stories about this company appear to be from around 2021, coinciding with their most recent funding round. Of course, I expect that there are many companies out there working on sewing automation so it's not as if the whole industry is resting on this one company's shoulders. But, it was the one that was identified as a leader.

Oddly, the SoftWear website does not seem to have any videos of their Sewbots in action. I found a few snippets of their machines on YT but they were mostly too brief to really make out the capabilities of their machines (e.g.: like this video). This video is the most expositive one I could find. I will note that the sewing task that the machine is doing in that video is extremely straightforward (emphasis on "straight" lol) and the material being sewn seems relatively heavy and rigid compared to, say, typical t-shirt material. While it's entirely possible that their machines can do more than what's shown in that video, I think the true demonstration of automated shirt manufacturing is when we can see a machine that can sew collars and sleeves on thin, stretchy cotton-poly blends at scale.
posted by mhum at 2:42 PM on March 14 [5 favorites]

Comparing a robotic system to a human, the current labor force is lower skilled, lower ability, and a much higher attrition rate.

No.

I think you may be reading this opposite of how it was intended to be read. I believe it is meant to be a description of the current robotic “labor force.”
posted by atoxyl at 4:22 PM on March 14

Or actually, no, never mind, maybe it is referring to the current labor force versus a hypothetical robotic future? I don’t know, it’s a very poorly written article.
posted by atoxyl at 4:24 PM on March 14 [1 favorite]

Thinking nationalization and profit distribution is a ban is your first problem

How does open source models become nationalized? Open source models I'd say are better than nationalization as right now a nationalized thing would be under the whims of 2 people who seem rather untrustworthy.

And profit re-distribution already happens, they are called taxes.

So not sure what is wanted with this nationalization idea or how one then profit distribution a government owned thing as i'd had thought 'round these parts no one was in favor of treating government run things as profit centers what with government not being a business.
posted by rough ashlar at 5:25 PM on March 14

In general I think “figure out how to get AI and robots to do stuff so we don’t have to” is a bad problem to try to solve and leads to bad places.
posted by caviar2d2 at 5:59 PM on March 14 [1 favorite]