AI driven labor crisis?

I installed my first home AC system back in about 87. I did my second bigger one in 2001. I'm going through another AC upgrade now, and getting some help, cuz I'm getting old and cranky. It's amazing how common the three jobs are.

And yet, there is no way, no how any AI driven robot could do the job. The human experience in design, selection of materials and tools, fitment, installation, and testing is just amazing watching an AC pro do these thousands of complex task in the right order, with the right tools, and the right materials. I'm the amateur helping out this time, but the job is the same I did 40 years ago.

Humans are remarkable taskmasters.
 
I had a task at work, where we needed to open over 25000 cad files, press escape several times (to clear some dialog boxes), press several shortcut button combinations (to orient the view and turn on layers), take a screenshot, trim the screenshot, and save the image. Definitely doing this manually was out of question. Several quotes that I got to write this script were close to $5000 and some even around $10k. Free version of AI wrote this program within a minute, I spent another 2-3 hours making some edits, testing it and by the end of the day I had a working program. Ran it before I left and in the morning everything was done. So even with my limited knowledge of programming (only hobbyists level) I was able to pull this off. That's $5k-$10k that somebody is not getting. Few years ago the company would have definitely paid that and even more.
 
Barrett50 said:
And yet, there is no way, no how any AI driven robot could do the job.
Click to expand...
Actually, a properly developed AI could match the most experienced human out there... but is there a robotic system that could actually implement the required movements. There may be, but that is the big question.

I love some of the AI answers I've seen for various things that are just SO SO WRONG. But if you task AI with something it has been "trained" to do, it can be amazing. I don't expect AI to take over jobs like plumbing anytime soon just because the robotics part would probably be pretty expensive, but there will come a time in the not so distant future.

I see more administrative or clerical jobs going away. Anything that is basic data entry and then reviewing that data. Accounting is probably going to high on the list. Sure, some accountants "play games" with the numbers, but are they really? If there smart and they know the laws, then maybe they are just taking advantage of those laws. AI can easily learn ALL the laws and maybe take advantage of some a human accountant missed or, because s/he is human, forgot about.
 
M20K said:
but is there a robotic system that could actually implement the required movements.
Click to expand...
While some of it is CGI, he has a few "making of" videos on his channel showing the actual robots. And this was 8 years ago.

 
M20K said:
I love some of the AI answers I've seen for various things that are just SO SO WRONG. But if you task AI with something it has been "trained" to do, it can be amazing. I don't expect AI to take over jobs like plumbing anytime soon just because the robotics part would probably be pretty expensive, but there will come a time in the not so distant future.

I see more administrative or clerical jobs going away. Anything that is basic data entry and then reviewing that data. Accounting is probably going to high on the list. Sure, some accountants "play games" with the numbers, but are they really? If there smart and they know the laws, then maybe they are just taking advantage of those laws. AI can easily learn ALL the laws and maybe take advantage of some a human accountant missed or, because s/he is human, forgot about.
Click to expand...
I have some ham radio logbook information from long ago that I have been slowly, tediously, transcribing into the format that lets me upload it to online logbook systems. Handwritten pages from the early to mid 1990s. I decided to let ChatGPT have a crack at it. Snap a phone pic, upload, let GPT transcribe, then marvel at how wrong it was. I fed it corrections to the first page - it got about half the callsigns wrong - then we tried another page. That was much improved; it's learning my handwriting. We'll see how well it does with the third page. I think by the fourth or fifth I'll be able to just check for errors rather than making corrections.

This kind of thing really wasn't available just a very short time ago when I was managing a team of employees. If it were, it could have saved me a ton of time keeping track of projects and tasks, providing a lot of the data and boilerplate for performance appraisals, all kinds of things. In fact, I think an AI agent could have saved all of us a lot of time, especially time spent in meetings that produced no actual work but were nonetheless necessary and required.

If I were in the same situation with the same team today, don't think an AI agent would be able to replace me or anyone on my team - but it could make my job, and the jobs of the people working for me, a lot more efficient and productive.
 
I did not (and do not) use Siri because I did not think it was a good deal to give away the milk (training data) for free. I have the same opinion of LLM-as-a-service. Asymmetric information makes the world go round, I suppose. I'd much rather be a trailing-edge person and run a local model where I can control the goesintas and goesouttas.

I also wonder if there's a good analogy in additive manufacturing; the tech continues to improve, the number of applications will increase, but it's not going to roll out everywhere overnight; it just becomes another tool in the bag, to be used when appropriate, and the noobs will lose some fingers when they use power tools they are not yet ready to wield.
 
Bob Noel said:
Given that resources aren't infinite, how will, in this utopian paradise being discussed, resources be allocated?
Click to expand...
In the utopian paradise, resources are infinite because robot slaves produce them in infinite quantities.

I've also heard UBI tossed around without thinking about where the funds come from if nobody is working.
 
Barrett50 said:
And yet, there is no way, no how any AI driven robot could do the job. The human experience in design, selection of materials and tools, fitment, installation, and testing is just amazing watching an AC pro do these thousands of complex task in the right order, with the right tools, and the right materials. I'm the amateur helping out this time, but the job is the same I did 40 years ago.
Click to expand...
I've already argued above that I think plumbers will be among the first trades to be effectively replaced by AI and HVAC is subset of plumbing, so you would correctly guess that I don't think HVAC is some crazy magic AI won't be able to master pretty easily.

It comes down to this. Hopefully you read the Mythos link I shared above. One of the reasons, even today, something like 90 minutes into the AI revolution and with the bugs and hallucinations that exist today, it was successful in finding things that security experts and developers had missed (in some cases for a decade or more) is because it is way better at doing "thousands of complex task in the right order, with the right tools, and the right materials" than humans. Its attention doesn't waver. In the HVAC example, it won't shrug its shoulders and decide to substitute the part in its truck for the correct part to save a trip. It won't forget an important part of the service manual. It won't try to sell you a new system rather than replace a part, something the HVAC industry is pretty legendary for today. Think of it like this. Humans can build up experience and develop skills. A human with that advantage is a better HVAC tech and one without it. But humans also forget things and will run across HVAC systems they aren't familiar with to begin with. For AI, every site visit is an open book test because they will have access to everything, all the time, instantaneously.

35 years ago I did a project for Whirlpool to digitize all their service manuals and make them available to field support on location because it's unreasonable to expect a human to know everything about every product they have made. Today, that information still exists and, of course, anything manufactured in the past 20 years was digital first to begin with. All of that information is available to the AI and, again, humans generally have notoriously poor recall for specifics and details. This is part of the reason why it's expected to have a punch list of issues before accepting the keys and signing off on a new house, for example.

Even today, AI is pretty good at problem solving and process planning. When I wrote my previous note about plumbers I gave it a photo of my bathroom faucet and the inside of the cabinet/bottom of the faucet. It generated a 100% accurate list of the steps necessary to replace it. This list included the tools necessary, steps in the correct order and things to look out for. It did this based on one good photo (exterior) and one 10% blurry photo (interior, low light). I can vouch for the accuracy of its plan (including troubleshooting steps) because it was a faucet I had recently replaced. It even noted the water marks inside the cabinet that were the result of the leak that motivated me to replace the faucet a few weeks ago.

When reading your note, I asked it to give me a project plan for replacing my mini split and it successfully identified the steps, the tools and the materials needed. The AI part is already there. What doesn't exist is the robots to do the work. But, again, that's not a technological problem that's a money problem. The technology needed to build those robots exists today and someone pointed out that Elon Musk is moving aggressively in that direction.

Another example of problem solving. I've been collaborating with a company lately that designs enclosures. Today they are focused on small things that can be printed or CNC milled, but the business plan extends to other kinds of fabrication. I fed their AI the spec sheet for a single board computer I've been playing with lately and asked it to design me a project box. It read the 12 page document (attached), figured out what parts were relevant, ignored the watermarks, asked me about power requirements, noticed that the board has a handful of pins available for external circuits, asked me about my intentions and finally designed a project box that was customized for my project and SBC. Which I then printed.

I would argue this is at least as complex as the HVAC example. Maybe a little more so in the aggregate. Some parts are more complex, some less. But at least in the same ballpark. And it is already working.

I'll leave the economic and cultural implications for the philosophers, but it's hard to deny the reality of the technology.
 

Attachments

Last edited:
bflynn said:
In the utopian paradise, resources are infinite because robot slaves produce them in infinite quantities.

I've also heard UBI tossed around without thinking about where the funds come from if nobody is working.
Click to expand...
There are a million questions about how this will work, and you question is one of the good ones.
IMHO, there will always be a need for SOME people to work. These will be the ones with the ambition/greed to succeed. The rest of the people will accept some form of UBI, and thus the reason that I earlier stated that the division between the HAVES/HAVE-NOTS will become even greater than it is now. There will be a bunch of Elons (Trillionaires or higher) but there will be Billions of people on basic income, and not much in between, unless we find a way to get people to be productive when they don't have to be.

Something will be happening soon, but I doubt it will be in the few years that I personally have left.
 
Coincidentally, a very good opinion piece in the NYT this morning about this very subject. It's not rosy. Much of it predicts the increase in inequality that I talked about last night in this thread.

"Whether you talk with engineers, venture capitalists, founders or managers, or with doomers, accelerationists, lefties or libertarians, the so-called San Francisco consensus on the impact of A.I. for workers is bleak. Many are convinced that advanced A.I. will soon surpass human capabilities. This would produce tremendous growth and scientific achievement, but it would also displace millions of jobs as fewer humans are needed to make the economy run. The technology will depress economic mobility and exacerbate inequality, while ferrying power and wealth to the A.I. companies and the existing owners of capital."
 
bflynn said:
In the utopian paradise, resources are infinite because robot slaves produce them in infinite quantities.

I've also heard UBI tossed around without thinking about where the funds come from if nobody is working.
Click to expand...
But raw materials are not infinite even if robots gather/mine/process them. So how do we (as a whole) allocate them?

And UBI would have to come from taxing the relatively (compared to the population as a whole) few people/companies that are controlling the means of production.

But where will the companies get their profits/income to be taxed on in nobody has money to spend on these cheaply manufactured products?

John
 
A healthy economy needs consumers. No income means no consumption. You need the masses to be able to buy stuff, otherwise the whole castle will collapse.

Edit: jsstevens beat me to it.
 
It's absolutely astonishing how many people are so doctrinaire about the current economy that they can't imagine there might be a different one tomorrow with the spoils of production distributed in a completely different way.

Hint: There will be. Our modern economy, defined as the beginning of money, is something like 5000 years old. It's more like 2000 or 3000 years old if defined by the beginning of democracy. Or 200 years since the start of industrialization. It's a flash in the pan compared to the 100,000 to 300,000 year history of humanity. As a species we've had multiple economies. Everyone on PoA today just has the good luck (hopefully) of being around to see the next shift.

Maybe, in the words of Perry Farrell, "We'll make great pets"

1777566504870.png
 
Last edited:
It turns out that money is a quite flexible way of determining how each of us chooses what's important to us. I am quite aware that much of the world's population are deeply under resourced with our current money and that the distribution of it is badly skewed. But as a way to determine 1) how much is my time pursuing various things worth (I traded my time and skills for money for more than 40 years, as have many of us), 2 how important to me are a bigger/nicer dwelling? transportation? Food? Medicine (and differentiating between life, health and optional "upgrades")? It's a much more flexible way to do so than barter. And a much more personally controllable way than any top down allocation we've tried so far. It's not good, nor ultimately sustainable, that people are having to choose between things that should be universal (food vs. medicine vs. housing). But that is not the fault of money as a concept.

I say this knowing that for many years I had a sentence at the top of my resume that I really thought was a good philosophy: "Machines should work, people should think." I've now been around long enough to know that there are many great people-honorable, honest, caring, etc.-who really are not capable of making a job out of thinking. I don't value them less, but our economy does. And AI expands that to include many (if not all) people who can build a career out of thinking. (I include arts in this broad "thinking" category.)

What's the solution? I don't know. I hope it doesn't involve massive bloodshed before we figure it out.

I do know that so far, the people who are saying we just don't understand what the next system will be (at least on this thread) haven't proposed what it will look like.
 
DaleB said:
I have some ham radio logbook information from long ago that I have been slowly, tediously, transcribing into the format that lets me upload it to online logbook systems. Handwritten pages from the early to mid 1990s. I decided to let ChatGPT have a crack at it. Snap a phone pic, upload, let GPT transcribe, then marvel at how wrong it was. I fed it corrections to the first page - it got about half the callsigns wrong - then we tried another page. That was much improved; it's learning my handwriting. We'll see how well it does with the third page. I think by the fourth or fifth I'll be able to just check for errors rather than making corrections.
Click to expand...
Yep, this is what it can be really good at... and you're absolutely correct that you still MUST proof the results.
 
AV8R_87 said:
While some of it is CGI, he has a few "making of" videos on his channel showing the actual robots. And this was 8 years ago.
Click to expand...
There are a lot of great robotics out there. Don't know if I've seen one that can show up and fix ANY plumbing issue. Eventually there will be little bots that can do all sorts of things. And maybe 3 or 4 show up on a job and the one that is most suited is the one that does that repair. But for both actual devices and the cost to purchase and maintain them today just doesn't make it practical.

Probably the first robotic type device that will have a major impact will be the antonymous car. With the leaps in AI to better process all the sensors for a safe trip and be able to talk to passengers in multiple languages, etc., etc., I think we'll see massive leaps in the already growing industry. Uber drivers look out...
 
172andyou said:
There will be. Our modern economy, defined as the beginning of money, is something like 5000 years old.
Click to expand...
There was trade and items of value LONG before money. Pretty much the same system where if you want something you need to have something of value to exchange (cow or $100 bill, it doesn't mater). Issue will be when AI, robotics, automation or what ever non human starts doing the tasks. Then there will probably be a bigger divide of those that have and those that don't.

Then it will be interesting to see what "humanity" does to balance the playing field. Or do we go back to the dark ages for a few more centuries where you had the ruling class and those that may not actually be able to get enough food to live.
 
M20K said:
There was trade and items of value LONG before money. Pretty much the same system where if you want something you need to have something of value to exchange (cow or $100 bill, it doesn't mater).
Click to expand...
Yep, exactly. Drawing the line at money makes some sense for a "modern" economy. But yes, the entire point was that there was economic activity before that. If we're going to include barter, then several primates have economies. So do some kinds of bats (helping to feed unsuccessful feeders and building relationships and proto-culture as a result).

There are even crows and ravens that participate in human economics by that metric, bringing things to humans in return for food.

M20K said:
Issue will be when AI, robotics, automation or what ever non human starts doing the tasks. Then there will probably be a bigger divide of those that have and those that don't.

Then it will be interesting to see what "humanity" does to balance the playing field. Or do we go back to the dark ages for a few more centuries where you had the ruling class and those that may not actually be able to get enough food to live.
Click to expand...
Why would it be only for a few centuries? Why wouldn't it be forever?

And if it's a solvable problem, it would be great to solve it without a couple centuries of misery.
 
172andyou said:
So, a car that always does the opposite of what's expected/needed/desired?
Click to expand...
Sounds like the start of a good Douglas Adams book... a sentient but cantankerous AI automobile a la paranoid android.
 
M20K said:
Actually, a properly developed AI could match the most experienced human out there... but is there a robotic system that could actually implement the required movements. There may be, but that is the big question.

I love some of the AI answers I've seen for various things that are just SO SO WRONG. But if you task AI with something it has been "trained" to do, it can be amazing. I don't expect AI to take over jobs like plumbing anytime soon just because the robotics part would probably be pretty expensive, but there will come a time in the not so distant future.

I see more administrative or clerical jobs going away. Anything that is basic data entry and then reviewing that data. Accounting is probably going to high on the list. Sure, some accountants "play games" with the numbers, but are they really? If there smart and they know the laws, then maybe they are just taking advantage of those laws. AI can easily learn ALL the laws and maybe take advantage of some a human accountant missed or, because s/he is human, forgot about.
Click to expand...
Completely and utterly disagree. Recursive jobs(as noted prev) are fine. Putting the bumper cladding on a new car and repeat 3000 times is no problem for a robot, and it might even be able to learn to do a slightly different car. Going to 300 res homes, and upgrading the AC system is empirical knowledge that is far to varied to get right.
 
172andyou said:
Why would it be only for a few centuries? Why wouldn't it be forever?
Click to expand...
I think the reference to a pendulum swing is more integral with human nature. Eventually the masses by sheer numbers can make changes and often then a democracy of sorts form. As we can see in the world, there are still pockets of the old way, some pockets of the new way and possibly even some pockets trying to revert back to the old way.

So forever, I personally doubt it. But who knows for how long. A few more years/decades (relatively short) and things will swing back??? Or is there a decline and it does take centuries or forever as you're saying???
 
Barrett50 said:
Completely and utterly disagree.
Click to expand...
In the context of the discussion, you don't think an AI could be trained to fully understand and do the job of a plumber??? Remember, this is just the thinking part, I also said right now and for a while, there is no mechanical/robotic component that could be widely distributed to replace plumbers. (Sure someone may have a one-off prototype that can do it.)

But if properly trained (in the way you train AI) you're wrong that it couldn't out think 99.9% of the plumbers out there. Sure, there could be that .1% or let's even say that 1% or 5% that could match or beat the AI, but those are not the plumbers that usually show up at our door. Electrician recently showed up at a friends that was making some various upgrades and swapped out a switch to give him a dimmer. Didn't bother to check the other switch before leaving, so instead of being able to have both switches down for off and to be in alignment with the other switches, he had it backwards. My guess is an AI (or a better electrician) would not have done that.

Bottom line is that in a few years you're going to be shocked what AI can do.
 
M20K said:
In the context of the discussion, you don't think an AI could be trained to fully understand and do the job of a plumber??? Remember, this is just the thinking part, I also said right now and for a while, there is no mechanical/robotic component that could be widely distributed to replace plumbers. (Sure someone may have a one-off prototype that can do it.)

But if properly trained (in the way you train AI) you're wrong that it couldn't out think 99.9% of the plumbers out there. Sure, there could be that .1% or let's even say that 1% or 5% that could match or beat the AI, but those are not the plumbers that usually show up at our door. Electrician recently showed up at a friends that was making some various upgrades and swapped out a switch to give him a dimmer. Didn't bother to check the other switch before leaving, so instead of being able to have both switches down for off and to be in alignment with the other switches, he had it backwards. My guess is an AI (or a better electrician) would not have done that.

Bottom line is that in a few years you're going to be shocked what AI can do.
Click to expand...
I have no idea why you are asking me about AI doing plumbing. I never said or implied anything like that.

And - no, I won't be shocked. I've seen the hype before. It's a tool. Makes things faster. In 1942, the army hired hundreds of women to do Fourier Transforms, as a part of the Manhattan project. Eniac came along and did them in half the time, and put them out of business. Because - they were doing thousands of calculations per hour that could be automated.

Installing res AC systems in all kinds of dwellings with so many variables will not be part of the AI miracle(ahem).
 
Barrett50 said:
I have no idea why you are asking me about AI doing plumbing.
Click to expand...
Ummm... because you said you "Completely and utterly disagree." with a comment I made.

The comment you were disagree with was about plumbers. :cool:
 
jsstevens said:
But raw materials are not infinite even if robots gather/mine/process them. So how do we (as a whole) allocate them?
Click to expand...
Sorry, I meant to say that in the utopian paradise, resources are also infinite.
 
In this utopian society, I want a yacht. A really big yacht. Who decides if I can have one? Or does nobody have yachts, just small boats.
 
M20K said:
Yep, this is what it can be really good at... and you're absolutely correct that you still MUST proof the results.
Click to expand...
Unfortunately proofing results seems to be getting short shrift.
Barrett50 said:
Going to 300 res homes, and upgrading the AC system is empirical knowledge that is far to varied to get right.
Click to expand...
The result isn’t going to be robots able to fix everything, it’s going to be homes standardized so that robots can fix them. Cookie cutter cities designed for robots, not humans.
 
Hang 4 said:
In this utopian society, I want a yacht. A really big yacht. Who decides if I can have one? Or does nobody have yachts, just small boats.
Click to expand...

Canoes. Dugout canoes. As soon as someone invents a robot to make them.
 
Barrett50 said:
Installing res AC systems in all kinds of dwellings with so many variables will not be part of the AI miracle(ahem).
Click to expand...
HVAC is about a thousand times simpler than writing software and this thread started with tens of thousands of layoffs from the largest software companies in the in world. Just sayin'

Unfortunately proofing results seems to be getting short shrift.
MauleSkinner said:
The result isn’t going to be robots able to fix everything, it’s going to be homes standardized so that robots can fix them. Cookie cutter cities designed for robots, not humans.
Click to expand...
Nope. Not a chance. Where AI really shines is that things don't have to be standardized. You can get exactly the home you want because you aren't constrained by the efficiency needed for mass production. Mass production was a primitive way of offloading human labor by making processes repeatable and targets for automation with unthinking mechanical devices. With AI robots you'll be able to dictate everything about the new sunroom you want to add to your house and have it built completely from scratch. You won't be limited to only the options available as a kit from Sears.

And the "far too varied" argument falls down on another axis as well. Remember that the entire AI HVAC fleet learns the lessons of the others. So when one encounters a challenge on a 17th century restoration, the entire fleet learns the solution that worked from an aesthetic, physics, materials and legal (because old houses are often protected in some fashion) and has that information at their disposal going forward.

Look at the software example again. There is nothing cookie cutter about it and today's AI is able to outperform humans on exactly the kinds of tasks that require empirical knowledge and infinite variation. There isn't a human alive that can write code in every computer language ever created, figure out a custom deployment plan for the server infrastructure the company owns, analyze the potential security threats and document everything to create a record. AI can do that today.

Yes, today it still makes objective mistakes (as do human programmers and human HVAC techs), though anyone who has been doing this stuff for more than six months is gobsmacked at how quickly things are improving. From an AI perspective, if it can do software at scale, it can do pretty much anything. All that's missing is sufficient robotics to allow it to work in the physical world. Just yesterday, AI beat medical doctors at making accurate diagnosis.

I really just can't understand why anyone would choose HVAC as the hill to die on. It's easy compared to challenges that have already been met.
 
Last edited:
172andyou said:
So when one encounters a challenge on a 17th century restoration, the entire fleet learns the solution that worked from an aesthetic, physics, materials and legal (because old houses are often protected in some fashion) and has that information at their di
Click to expand...
Assuming that the one can figure it out.
 
MauleSkinner said:
Assuming that the one can figure it out.
Click to expand...
Yep. Given that they have access to everything known about 17th century building practices and an HVAC tech doesn't, I feel very comfortable about making that prediction.
 
172andyou said:
Yep. Given that they have access to everything known about 17th century building practices and an HVAC tech doesn't, I feel very comfortable about making that prediction.
Click to expand...
I’m sure they’d do just fine replacing a 17th century HVAC system when the compressor fails.
 
MauleSkinner said:
Exactly. They’ve largely quit building homes for the occupants already. They’re built for the convenience of the builder. If robots start building and maintaining homes, why wouldn’t they build them for the convenience of robots?
Click to expand...
They may, in other ways. But cookie cutter isn't a convenience that matters to a robot. They are equally capable of doing a new design for every single house or the same design a hundred times. My prediction is that the humans will prefer the former, so that's what will end up happening.

MauleSkinner said:
I’m sure they’d do just fine replacing a 17th century HVAC system when the compressor fails.
Click to expand...
Obviously I meant installing a new system into a crusty old house that had never had one, but you did make me chuckle.
 
So - the AI future all society is going to be commoditized? And all of the next generations are going to go along with that? To fit the mantra of AI's taking over all labor.

Yeah - sure.

 
You must log in or register to reply here.
Back
Top