'AI' to hit hardest in U.S. heartland and among less-skilled: study

[Checkerboard Strangler]

Those are two awesome examples that are both wonderfully counterintuitive.


Food prep is a good one. There isn’t a lot of buzz behind it yet, but it makes a lot of sense. Fast food jobs don’t pay enough to provide a living wage and the product doesn’t support the prices required to provide a living wage. The preparation is also simple motions that are very consistent and well defined. I could see a Big Mac being made in a ford assembly line today. The real effort would be getting the costs down and reliability up.... and the hmi for the customer. You’re not going to need a ton of additional sensing or adaptation beyond traditional automation to get there.


But even then, the economics are non trivial. Automating an entire kitchen is probably cost prohibitive. Hardware alone would run you million(s?) plus integration and maintenance. So I’d first look to small single manipulator compact units that are more akin to a fast food truck or kiosk. I’d say you’d need 10-50 units before the whole thing becomes revenue neutral (though that’s just a back of the envelope)


House cleaning is a completely different story. That will be one of the last things to be automated, if ever. At this point no one is even remotely close. The problem is orders of magnitude more complex than flipping burgers. Manipulation is extremely difficult, especially when you’re dealing with objects that deform. Try to imagine computing a mathematical model of a shirt that reacts to picking it up and moving it, and then try to morph that model that onto a shirt laying in a pile of clothes so you can predict where you need to grasp it to begin folding.


Economically, even though you’re also dealing with lower paying jobs it’s totally different from a fast food setting. Instead of many consumers interacting with a single robot, you’re talking about robots as an appliance. Even with the rosiest glasses you’re almost certainly looking at 100k+/robot if it were even possible. That’s too much for a single consumer.

Obviously robot vacuums are here now, but improving them is going to be tough. Cordless vacuums aren’t all that great and dragging around a tether (cord) is a huge complexity. And there’s also a paradox here, as the more capable you make a robot the more susceptible it is to exhibiting adverse emergent behaviors.

---You're staring too closely at the individual letters instead of the paragraphs.

1. Hotel room cleaning:

a) First string room supers go through, pick up shirts and other sundries, toiletry stuff, place them out of the way.
b) Robotic cleaning machines sweep through, vacuum, make beds, clean bathrooms,
c) second string goes through, resets rooms.

Minor room design changes make it work. Each room has a no man's land where suitcases, sundries etc can be put during cleaning, same with bathrooms. The same software that makes supermarket self checkout work can be adapted. Remember, we are talking about in ten years, not tomorrow. You can bet this is already being researched.

2. Food prep:

a) Again we are not necessarily talking full automation. A robot chef cannot poach sixteen eggs while basting seven others and make German potato pancakes all at the same time during breakfast rush however a tightly organized crew of helpers can feed the necessary ingredients, start the bots on a task and then come back through and finish.

b) Bots can also do the sidework, prepping for instance. Any bot can par-boil a pot of potatoes and then peel 300 pounds of them for the day and also prep corn, open cans, mix batters, make sauces and gravies, flip steaks and pancakes.
Some kitchen redesign will be required.

Again, we are talking a decade down the road or thereabouts so economies of scale are in play. It's not tomorrow.

---You present incredibly discerning questions, the kind which industry researchers are no doubt chewing on as we read this.
Now let's do warehousing, which thanks to Amazon and Walmart, are already well on the way to automation.

 

[jmotivator]

You are absolutely right, and they're going to have to become a LOT MORE AFFORDABLE for the typical mope to afford them.
We have a task before us. We have to be ready and able to train an entire generation coming up with these new skills, and retrain those who will be displaced as well.

If you want colleges to be affordable then get the government out of student loans. The current law to pay off those loans is something like "10% of income for 10 years and then the government eats the remainder". That is a blank check to colleges that dislocates the cost of school and the value of the resulting education.

Trade schools are very affordable, and many allow you to earn the education through apprenticeship. Trade schools are probably a better target for blue collar workers anyway.

This is just not an excuse for yet another massive government program.

 

[Checkerboard Strangler]

If you want colleges to be affordable then get the government out of student loans. The current law to pay off those loans is something like "10% of income for 10 years and then the government eats the remainder". That is a blank check to colleges that dislocates the cost of school and the value of the resulting education.

Trade schools are very affordable, and many allow you to earn the education through apprenticeship. Trade schools are probably a better target for blue collar workers anyway.

This is just not an excuse for yet another massive government program.

Have you ever attended any trade schools? Which ones, when and what did they cost you?

 

[Mithros]

---You're staring too closely at the individual letters instead of the paragraphs.

1. Hotel room cleaning:

a) First string room supers go through, pick up shirts and other sundries, toiletry stuff, place them out of the way.
b) Robotic cleaning machines sweep through, vacuum, make beds, clean bathrooms,
c) second string goes through, resets rooms.

Minor room design changes make it work. Each room has a no man's land where suitcases, sundries etc can be put during cleaning, same with bathrooms. The same software that makes supermarket self checkout work can be adapted. Remember, we are talking about in ten years, not tomorrow. You can bet this is already being researched.

2. Food prep:

a) Again we are not necessarily talking full automation. A robot chef cannot poach sixteen eggs while basting seven others and make German potato pancakes all at the same time during breakfast rush however a tightly organized crew of helpers can feed the necessary ingredients, start the bots on a task and then come back through and finish.

b) Bots can also do the sidework, prepping for instance. Any bot can par-boil a pot of potatoes and then peel 300 pounds of them for the day and also prep corn, open cans, mix batters, make sauces and gravies, flip steaks and pancakes.
Some kitchen redesign will be required.

Again, we are talking a decade down the road or thereabouts so economies of scale are in play. It's not tomorrow.

---You present incredibly discerning questions, the kind which industry researchers are no doubt chewing on as we read this.
Now let's do warehousing, which thanks to Amazon and Walmart, are already well on the way to automation.

Thats a good distinction, at least at a high level. Robots augmenting workers rather than replacing them.

That probably doesn’t work in hotels though. Its augmenting the easiest part of low wage workers jobs while not actually removing them. Instead of a single low paid maid cleaning a room, now you have two maids, a robot, and at least one person to operate and repair the robot. It’s only cost effective when the robots remove enough people to pay for themselves.

And bed making is incredibly tough. Here are two examples.
https://m.youtube.com/watch?v=8HOXXygcgOM
https://m.youtube.com/watch?v=dqiHo
The first might look impressive, but is completely faked. The cover is “messed up” in a very particular way. It’s just running a preprogrammed script. It was probably tougher to move the bed into the cell than make the video. The second is more akin to reality, but even that is slow and likely just following heuristics. (Find the corner of the sheet, find the corner of the bed, compute a path to grasp the corner of the sheet, move it to the corner of the bed, release..

In terms of cooking, doing all of those things simultaneously is actually much easier for a robot than a human. Simple independent tasks scale well, compound interdependent tasks don’t. I know the temperature of the grill so I can estimate and schedule cooking times. I know the grill location relative to the robot so I can preplan every motion. I can monitor as segregate my workspace to manage computational complexity. I can control lighting so that vision classifiers can be trained to know when food is done approaching the accuracy of a human. It would still take a lot of work, but it’s all tractable.

And if you can make the economics work you’ve created a bunch of high paying jobs to replace jobs that don’t support a living wage. Just like the economy doesn’t support an army of people to dig ditches, but does support heavy equipment operators and mechanics with well paying jobs to perform the same tasks.

Warehousing has gone down a bit of a dark road thanks to amazon. Amazon centers are kind of fully automated already... they just automate it with humans instead of robots... which imo is pretty horrible. It’s like they’ve solved the distribution problem, just not the manipulation one.. so they task the humans as if they’re robots. Amazon has sponsored a few pick and place challenges that a bunch of people responded to. There’s a bit of progress, but it’s still an incredibly hard problem. Maybe they get there.. maybe not.

Also, the economics of developing robots has changed drastically thanks to self driving cars. A lot of the talent had been concentrated there and salaries have skyrocketed.

Backing up... all of this is a good thing. Productivity gains have been flat at 1%. Same with population. If we don’t change those numbers we’re stuck at 2% growth per year. Also we’re looking at changes that happen on a scale of decades. It’s disruptive, but we can adapt. I’d even argue that the pain we’ve already felt is worse than what is to come.

What isn’t going to happen is an AI apocalypse where a few people will produce Swiss army robots that will solve everyone’s problems through deep learning.

 

[Xelor]

Job Training for what?

Seriously? Did you even look at the rubric article?

Job training to perform some sort of work that (1) is less susceptible to being replaced by machinery and that provides one with a decent income.

• Four fundamentals of workplace automation
• Jobs lost, jobs gained: What the future of work will mean for jobs, skills, and wages
• The Risk of Automation for Jobs in OECD Countries
• THE FUTURE OF EMPLOYMENT: HOW SUSCEPTIBLE ARE JOBS TO COMPUTERISATION?
• The computerisation of European jobs

In other words, training to do something at the bottom of the list on the right side of the image below. A good place to start is with anti-indolence training.

I am playing devils advocate. I know the free market is the way out especially the FULL embracement of it. But are people prepared to do that? I suspect most are not. Not really. It requires an entrepreneurs mindset. Thats something that, in the USA is sadly not as prevalent as perhaps is ought to.

For me its owning the machines, and investing other companies and individuals with machines, and have a finger in many pies. Thats how I thrive in an automated economy.

The op presented training as an option to relieve displaced workers. I ask train them for what? If you train them for another job then they will be displaced again. There in lies the rub.

I also suggest that the article you presented, paints a rosy rather optimistic picture. I suspect reality will lie somewhere less optimistic.

Red:
If they're not, well, that'd be among the types of training they need, now wouldn't it....

 

[Mithros]

If you want colleges to be affordable then get the government out of student loans. The current law to pay off those loans is something like "10% of income for 10 years and then the government eats the remainder". That is a blank check to colleges that dislocates the cost of school and the value of the resulting education.

Trade schools are very affordable, and many allow you to earn the education through apprenticeship. Trade schools are probably a better target for blue collar workers anyway.

This is just not an excuse for yet another massive government program.

Another way to think about that is that the government gets ~21% of all income. College grads make 56% more than non college grads. ie the government will pull in at least 10% more in revenue per year for every college graduate they generate. Average student loan debt is 37k which is roughly median income. Conservatively the government would break even eating student loan debt in less than 10 years under the current tax structure and would make ~100k over the life of an average person.

 

[RetiredUSN]

For starters, how about learning to fix the machines that just took your job away?

AI will not kill the job market in the midwest.

AI will never eliminate trucking jobs.

 

[Helix]

'AI' to hit hardest in U.S. heartland and among less-skilled: study

this is incomplete. i have a graduate degree, almost two decades of experience, and i work in a highly skilled field. they are also trying to automate my job. i watch their progress daily, and though it's not going that well so far, it will probably happen eventually. this is the case for many if not most jobs.

 

[Mithros]

Seriously? Did you even look at the rubric article?

Job training to perform some sort of work that (1) is less susceptible to being replaced by machinery and that provides one with a decent income.

• Four fundamentals of workplace automation
• Jobs lost, jobs gained: What the future of work will mean for jobs, skills, and wages
• The Risk of Automation for Jobs in OECD Countries
• THE FUTURE OF EMPLOYMENT: HOW SUSCEPTIBLE ARE JOBS TO COMPUTERISATION?
• The computerisation of European jobs

In other words, training to do something at the bottom of the list on the right side of the image below. A good place to start is with anti-indolence training.



Red:
If they're not, well, that'd be among the types of training they need, now wouldn't it....

That’s a good post. One caveat, I’ve suspected that this graph was made by a business person.... and they confirmation biased themselves way down on the list.

My advice would be to first go into STEM, especially the E. If that’s not a good fit, pick up a trade.

 

[Mithros]

AI will not kill the job market in the midwest.

AI will never eliminate trucking jobs.

If you’re talking local trucking I’m with you. But long haul trucking is where I’d first expect to see large scale deployment of driverless vehicles.

Even if that doesn’t happen, fuel issues are going to kill long haul trucking. Oil isn’t forever and a national high speed rail network makes more sense imo.

Basically, if you’re a trucker today you’re probably safe. But if you’re starting out, pick something else.

 

[Deuce]

You might want to read this: AI Will Create Millions More Jobs Than It Will Destroy. Here’s How
But again, free market, right?

This article has no justification for the claim other than "it hasn't happened yet."

 

[RetiredUSN]

If you’re talking local trucking I’m with you. But long haul trucking is where I’d first expect to see large scale deployment of driverless vehicles.

Even if that doesn’t happen, fuel issues are going to kill long haul trucking. Oil isn’t forever and a national high speed rail network makes more sense imo.

Basically, if you’re a trucker today you’re probably safe. But if you’re starting out, pick something else.

The first run away AI truck to kill a bus load of kids will bring a major halt to AI trucks sharing busy roads..

 

[Deuce]

The first run away AI truck to kill a bus load of kids will bring a major halt to AI trucks sharing busy roads..

Much like the first human-driven truck did?

 

[jmotivator]

Have you ever attended any trade schools? Which ones, when and what did they cost you?

I went to a trade school to learn IT back in the 90s. It cost me around $8,000. It's how I made it out of social work.

Trade schools today run closer to $30k, which is still far cheaper than the average bachelor's degree and you end up with a marketable skill unlike the majority of bachelor's degrees.

 

[What if...?]

Someone has to design them, program them, supervise the factories where they are built, etc. Unless you're reading a science fiction novel, you can never eliminate humans from the equation. But the humans involved here will need the required training & resulting skill set.

So that's a few people. Not billions.

The real problem is the worlds economy is financing a global game of Highlander.

Too many with too strong an urge to empire. All trying to win the world.

They don't care about you, consider you expendable commodities. Expensive ones at that.

Capitalism has sought to eliminate labor where it can and keep wages down where it can't.

"Job creator" is a PR term. Many businesses today use minimal labor to make billions. None hire people out of the kindness of their hearts. Pensions, job security, benefits all gone or on their way out.

So as AI and robots become more sophisticated capitalism will use them as soon as they cost less than people. And use the dividend to design more machines to do more jobs instead of people.

The type of kind that will thrive in the sunset of labor is pretty rare. Technical, creative.

No idea what the average and below have to look forward to.

 

[Checkerboard Strangler]

I went to a trade school to learn IT back in the 90s. It cost me around $8,000. It's how I made it out of social work.

Trade schools today run closer to $30k, which is still far cheaper than the average bachelor's degree and you end up with a marketable skill unlike the majority of bachelor's degrees.

Sorry, unaffordable unless you consider that kind of debt to be affordable on a 25 thousand dollar a year income, which is what most applicants would be.

And thanks to recent deregulation, said trade schools are ONCE AGAIN free to be diploma mills that don't teach you anything useful, and you're still on the hook for life. We had started to make progress toward some standards and some recourse for people who got burned but 45's administration reversed most if not all of it.

 

[What if...?]

New projects and initiatives that were formally not financially feasible will come within reach. The only people who won't be employable are those who can't or won't learn.

How many people do you know who aren't creative, won't ever design software regardless of how hard they try to learn? You surely know some.

Multiply that by seven billion, remove "duplicates". How many billions will have no place in the "new order"?

 

[jmotivator]

Another way to think about that is that the government gets ~21% of all income. College grads make 56% more than non college grads. ie the government will pull in at least 10% more in revenue per year for every college graduate they generate. Average student loan debt is 37k which is roughly median income. Conservatively the government would break even eating student loan debt in less than 10 years under the current tax structure and would make ~100k over the life of an average person.

"Roughly media income"? I think you are off by about 62%.

Also, do you have a link to the study you are using for income? I've seen studies that put the income at about $50k for a college graduate, but they are not actually surveying graduates, or even a random sampling of college grads, just employers and select degree categories.

Plus, there appears to be a correlation between loan participants and the federal government running out of money. But that is all beside my point.

I was not arguing that the federal government is losing money, I am arguing that the program is separating cost from value in higher education, which it is. If two students gets a degree in, say, social work, one from a school that charges $40,000 versus one from a school that charges $90,000, they will both end up paying the same for their education assuming the same pay for the same job, so there is no incentive for colleges to price competitively, nor for students to shop competitively since the the actual realized debt payments per discipline for a degree is normalized.

 

[Checkerboard Strangler]

So that's a few people. Not billions.

The real problem is the worlds economy is financing a global game of Highlander.

Too many with too strong an urge to empire. All trying to win the world.

They don't care about you, consider you expendable commodities. Expensive ones at that.

Capitalism has sought to eliminate labor where it can and keep wages down where it can't.

"Job creator" is a PR term. Many businesses today use minimal labor to make billions. None hire people out of the kindness of their hearts. Pensions, job security, benefits all gone or on their way out.

So as AI and robots become more sophisticated capitalism will use them as soon as they cost less than people. And use the dividend to design more machines to do more jobs instead of people.

The type of kind that will thrive in the sunset of labor is pretty rare. Technical, creative.

No idea what the average and below have to look forward to.

Highlander, Handmaid's Tale, Hunger Games, Game of Thrones, yup...
Hey, you know what? It doesn't matter whether you, me, Jmotivator or anyone else thinks we need to force the elites to contribute and invest or not.
It really doesn't, and it doesn't look like they're about to have any crisis of conscience about it either.
So it is popcorn time, as we watch our Hunger Games version of capitalism attempt to create an illusion big enough and shiny enough that people will buy it.

In the end, when it fails, as all illusions do, there will be blood. This time there will be a lot more blood than any other time in world history. You cannot fool all of the people all of the time, and when well over a third of them in this country alone are facing the prospect of being out of work, in debt, destitute and told they have no rights, that is when the blood will become an eruption.

Nobody wins one of those things.
We could save ourselves now and have a modified regulated capitalist system that serves the working class as their tool and still have plenty of rich people walking around enjoying their fortunes, with the kind of upward mobility and affordable cost of living we had in the New Deal, or the eruption could just wipe it all out and drown us for a couple of years and then we emerge fully bankrupt from coast to coast, and most likely under some kind of extremist government.
Flip a coin, because in the end it doesn't really matter if it's extreme right or extreme left, it won't be realistic or here to serve anyone except the elites once again.

I don't look forward to us being either 1930's Germany with bigger freeways or Cuba with larger shopping malls.
And I guarantee you the rest of the world isn't going to put up with our crap either no matter which way it goes.

 

[PirateMk1]

Seriously? Did you even look at the rubric article?

Job training to perform some sort of work that (1) is less susceptible to being replaced by machinery and that provides one with a decent income.

• Four fundamentals of workplace automation
• Jobs lost, jobs gained: What the future of work will mean for jobs, skills, and wages
• The Risk of Automation for Jobs in OECD Countries
• THE FUTURE OF EMPLOYMENT: HOW SUSCEPTIBLE ARE JOBS TO COMPUTERISATION?
• The computerisation of European jobs

In other words, training to do something at the bottom of the list on the right side of the image below. A good place to start is with anti-indolence training.



Red:
If they're not, well, that'd be among the types of training they need, now wouldn't it....

You have a point. I should have clarified. That the training they may need is not job training, but training in entrepreneurship, and business.

 

[What if...?]

And you think that's a smarter idea than investing in Joe Six Pack and his progeny?

Nice.

I don't think education will help much, honestly. Creativity is a rare trait. So is the ability to design complex systems.

All this technology was supposed to free man from labor. Lives of abundance free from toil. Letting us explore being human, foster our creativity, try new ways of living.

And there's no reason it can't. Except then all those guys trying their hardest to be the most wealthy and powerful would see their game end.

And I don't see them letting that happen. Because their drives are about status. How much higher on the ladder they are than everybody else. That takes a lot of money. Money that won't be around if AI makes scarcity a thing of the past.

 

[Checkerboard Strangler]

I went to a trade school to learn IT back in the 90s. It cost me around $8,000. It's how I made it out of social work.

Trade schools today run closer to $30k, which is still far cheaper than the average bachelor's degree and you end up with a marketable skill unlike the majority of bachelor's degrees.

PS: I also went to a trade school in the LATE 90's to learn IT, and it too cost me about seven thousand dollars.
It's how I tried to recover from losing $350 thousand dollars in the 1994 Northridge Quake.

And that school wound up on a list of schools under investigation for being a diploma mill, but they recently got a get out of jail free card because the entire case against all those schools has now been wiped out.

I wound up learning what I learned on the job instead.

 

[Checkerboard Strangler]

You have a point. I should have clarified. That the training they may need is not job training, but training in entrepreneurship, and business.

The entire population cannot all be entrepreneurs. What are you people even talking about?
The entire population of the USA all turning into day traders, insurance fund execs, CEO's and software designers?
Yeah, right.
We can sell insurance to each other.

 

[What if...?]

LOL No. They won't be able to fix themselves.

Most systems are modular. Module fails it sends a message to repairbot that simply runs over and replaces the module.

Make the modules modular and another robot can repair the bad module by replacing the failed card(s) inside. Cards maunfactured by robots.

 

[jmotivator]

Sorry, unaffordable unless you consider that kind of debt to be affordable on a 25 thousand dollar a year income, which is what most applicants would be.

I'm sorry, what trade are you talking about that only makes $25k a year? And do you ever expect a college education to be cheaper? than $30k? Add to the much lower cost of a trade school the fact that you are earning 2 years earlier than you would even get your college degree.

And thanks to recent deregulation, said trade schools are ONCE AGAIN free to be diploma mills that don't teach you anything useful, and you're still on the hook for life. We had started to make progress toward some standards and some recourse for people who got burned but 45's administration reversed most if not all of it.

Colleges have been turning out boat loads of useless degrees for years and years. Do you think making them cheaper would fix that?