When it comes to automation, your perception of what is dull and rules based is very different than what an automation engineer would consider easy to automate. Humans do things trivially that can be impossibly hard for a machine, and some things that are very hard for us are trivial for a machine.
For example, one of the most dull and rules based jobs is picking up trash on the street or in parks. Any four year old child can be taught to walk around and pick up loose garbage and put it in a bag. And yet, it would be incredibly hard to program a robot to do this in a general way, as trash often collects in bushes or under benches or other places that require agile grasping, terrain can be uneven, trash can be hard to identify from other stuff you don’t want to collect, it takes a lot of energy to stay mobile for long periods of time, etc.
Cleaning a house is a task that doesn’t require a lot of training, and which employs a lot of lower educated people. But we don’t have robots replacing Molly Maids, because it actually takes an amazing amount of decision-making, general dexterity and mobility to clean a house. You have to reach up to shelves, go up and down stairs, handle fragile objects, move furniture carefully without scratching things, and on and on. We don’t today have a robot that can even come CLOSE to replacing a house cleaner. And we don’t really know how to do any better at this point.
Robots ‘took over’ on assembly lines relatively quickly (as in, a few decades) because assembly line work is low-hanging fruit. Grab known part from belt A, transfer to Belt B. Pick up #34 bolt and insert it into hole that is always in a known place. This robots can do.
The next level of factory robots can use limited AI and machine vision to have a little extra flexibility, and can be used to spot things like paint discoloration and elongated holes, replacing some of the more rote Quality Assurance work.
These types of activities are far, far different than, say, installing plumbing in a house, repairing a damaged road, or even working in a factory that makes bespoke items which are not part of an extremely rigid and well defined process.
And then it turns out that you need people anyway, because you need feedback from real brains at the bottom to correct for mistakes made at the top. I work in factory automation, and I can tell you that every process meticulously designed by engineering teams gets modified when it hits the real world. Modified by humans working on the floor who have to work around the little things that were missed by the planners. Planners don’t have perfect information, and local knowledge is more important than you would think.
And getting all these machines to work together without interaction problems is extremely hard. An entire industry of analysts exists to shake out the bugs in automated assembly lines and keep them running efficiently. Because machines are dumb, and can’t handle the problems that can crop up when perfect designs meet an imperfect world.
The general failing of people who make sweeping claims about how some new technology is going to ‘take over’ is that problems look very different from 20,000 ft than they do when you are down in the weeds sweating the details. The Segway was going to solve the ‘last mile’ problem and finally cause us all to embrace mass transit. Cities would be changed forever. Remember those predictions? They never happened because an idea that looked simple and obvious in the broad strokes couldn’t survive the details. Flying cars have been ‘just around the corner’ for five or six decades now. It’s just those pesky details that have to be dealt with, then we’ll all be flying to work.
The ‘details’ still to be solved before robots are going to take away general human labor involve needing batteries that we have no idea how to build, a general intelligence AI that we have no idea how to build, building a mechanical mechanism that can do all the things a human can do, for as long as a human can do it, without requiring more effort in maintenance than a human wage, which we do not know how to do.
The general intelligence part is the real stickler here. To this day, we have NO IDEA how to make a computer that exhibits general intelligence and that can make judgements about activities that it has never before been exposed to. We do not know how to make machines that exhibit creativity, which can spot opportunities for improvement without being programmed, etc. Things humans do to add value to their work all the time. The flexibility of the human mind and body corrects for an awful lot of mistakes and oversights made by management, or for changes in working conditions that were not anticipated.
Let me give you a quick example: A process might be written to say, “You will receive a box of labels. Remove one label from the box at a time, and affix it to the product.” The box is supposed to be delivered to the bench beside you. But today, someone delivers the box and puts it across the room. You are proud of your new advanced AI robot that can use machine vision to find the box, no matter where it is. So the robot searches, finds the box… and proceeds to take one label at a time out of the box, carry it across the room, affix it to the product, then go back and get another… In the meantime, the human just picks up the whole box and puts it where it belongs. Then that human might inform someone that the box is being dropped in the wrong place so the process can be improved and the error corrected.
This may sound trivial, and if you knew this was the exact problem the robot would face you could obviously program it. But there are thousands of little mistakes and deviations from plan that happen all the time, and in ways unknown to everyone before they happen. it’s our workforces of creative, intelligent and flexible humans that keep everything running when stuff doesn’t work out as planned.
I’ve never worked in an office job where any metric was used to track my performance besides how many hours I bill to clients. I wouldn’t want to. I would find it incredibly depressing.
Which is not to say that a large part of my job isn’t breaking down processes into discrete tasks that are quantifiable.
I disagree with your analysis. I think it will become a problem much sooner than you think. It won’t wait until 50% are unemployed/unemployable before it’s a problem. I think once it’s over 12% it’ll become a problem. By the time it’s 18% it’s gonna be craziness.
It isn’t that the horse thing is relevant; it’s that it is analogous.
What new industries do you foresee being created that will require more human labor (and management) than computer and machine labor?
No, I don’t. Could you cite them? Are they predictions from outside observers or are they marketing blurbs from the manufacturer? Because that sounds like a marketing blurb.
Labor loses every time. The fact that humans overall see improved live isn’t what the statement is about.
Please list all the fields where machines (or other automation), once available, did not replace most human labor.
Farming? Weaving? Making clothes? Washing clothes? Making french fries from potatoes? Translating written languages? Printing things that people write so other people can read them?
Labor always loses to machines.
The fact that people start doing something new to them doesn’t refute that when it’s humans vs. machines, the humans always lose those jobs.
So what new industries do you see opening up that will require humans? That entail things that cannot be done by anything other than a human? And that humans will want to do?
LOL. Yes let’s put to one side whether it makes our lives better.
Yeah.
What idiot invented the plough? Didn’t (s)he realize she was taking away great jabs digging in the dirt with our bare hands? Now we sit around with ample food for everyone, but at what cost? At what cost!
That question has a surprisingly long history. It’s the nature of human progress that we can’t tell what the next step would be, but there always is one. You think the world right now is perfect?
Secondly the things that can only be done by humans is still pretty big as Hellestal has explained. It’s not that any given, very specific task can’t be automated. It’s that humans have general intelligence and be trained to do a millions of different tasks. Essentially humans are redundant only when you have “strong AI” and there’s no indication we’re getting any closer to that.
Um, every industry that still has humans working in it? Lots of jobs COULD be automated, but haven’t been because either the automation costs more than the people, or because the people have other values the robots don’t, or because the robots just aren’t as good as people in those applications.
For example, online food ordering or food ordering kiosks are easily done today, yet the vast majority of food vendors have actual people doing it. Why? Because other humans would rather interact with a person than a machine, and because until recently the people taking orders cost less to the organization than robots do. The biggest threat to those jobs in fast food is higher minimum wage laws, which push those worker’s prices to the point where machines can compete. But in restaurants we still generally have human servers. We could probably automate cooking, and in packaged food factories we do. But people like the variety and artistic flourishes of human cooks, and automating a small restaurant is too cost prohibitive.
For that matter, we also have buffets, which are like assembly lines for food that drastically cut down on the amount of labor required to prepare and serve food. So why aren’t all restaurants adopting that kind of model? Because people like other experiences - ones that require humans in the mix.
As I said, I work in factory automation. There are lots and lots of jobs in factories that could theoretically be automated but which aren’t because either the cost/benefit isn’t there, or because the factory just doesn’t want to give up the judgement of humans in certain cases, or just because the project would be way too expensive for the value gained.
See, you are cherry picking industries that HAVE been automated to make your point that ALL industries can be automated. This is your bias showing. When trying to gather evidence to determine the truth of a proposition, a more productive path to the truth is to try to think of counter-examples rather than examples that support your own biases.
For example, we have not and don’t know how to automate simple manual labor tasks such as installing house plumbing or electrics, installing furniture and carpeting, painting walls, etc. House construction is still largely a human activity.
There is very little automation in the health care industry. Nurses aren’t getting replaced any time soon.
Heavy construction uses very little in the way of robotics. Lots of labor saving machines have been around since the start of the industrial revolution, but they are dumb machines operated by people.
Since we have more labor opportunities now than at any time in human history, and workers get paid more than at any time in human history, that’s a very questionable statement to make.
What has actually happened is that machines have ‘taken over’ the worst kind of jobs for people - the jobs that involve heavy lifting or the kind of soul-killing assembly line jobs that people aren’t well suited for but machines are. These were also the jobs that paid the least. This has freed those people to move up the value food chain and earn more money doing things that are less boring or physically difficult. Or are you really going to argue that the loss of sweatshops is a loss for labor? If not, just what do you think causes sweatshops to go away? It isn’t political activism: It’s capital investment and machinery.
You could be as pro-labor as you want, and you still can’t make a job pay more if it simply doesn’t offer enough value to do so. To gain that value, you generally have to magnify the value of human labor. That’s a huge benefit to the laborer. A ditch digger with a shovel is way more productive, and can charge more for his labor, than one with a stick or with bare hands. And a ditch digger with a power backhoe can earn enough to make a 1st world living. All due to those horrible machines. And the guy with the backhoe can work while sitting down and moving a few levers, instead of moving thousands of pounds of dirt.
Human laborers are the big winners when it comes to machines. Why do you think salaries have increased steadily for hundreds of years? Why do you think an auto worker can make $40/hr, while a subsistence farmer who puts in much more labor makes nothing? Hint: It’s not because of labor unions, or political systems. It’s because auto worker’s labor is magnified by having access to millions of dollars of machines, and because the worker’s labor is magnified by existing in the context of an organization that can efficiently turn that labor into consumer value. The industrial revolution is the best thing to happen to workers. Ever. By far.
If we knew that, we could run out and get rich. But note that there is a gigantic industry of people out there now making reasonable livings providing podcasts, youtube channels, blogs, etc. Not a single one of those jobs existed twenty years ago, and thirty years ago you could never have predicted that ‘podcaster’ would be a thing.
Another huge industry has sprung up selling used goods, making bespoke art (etsy), making one-off items and 3D printing them, etc. There are people making a living now creating digital online content for games. Movable type ended manual scribing, and desktop publishing destroyed movable type. Technology marches on. All those scribes and typesetters lost their jobs, and if you’d asked then what kind of jobs would replace them, you couldn’t have predicted that there would be industries of web designers, for example.
30 years ago, no one had heard of web designers. Ten years ago, ‘drone operator’ wasn’t a thing - not even a predictable thing. Drones are an example of where a new form of automation has enabled entire new industries. Drones have also given individuals the ability to do sophisticated movie-making techniques that used to require helicopters, cranes, custom tracks and the like. This has democratised high quality content production, which helps expand the video industry and create new jobs. It sucks for the boom and crane operators, maybe, but they can retrain as drone operators and in the meantime the lower cost of access expands the entire market for everyone.
Automation was required to enable a rocket to land propulsively on its own landing pad after launching a payload. This has the potential to cut the cost to orbit by 50-90%. What markets will open up as a result? We don’t know, but it seems likely that there will be new jobs in a larger space industry in the future.
3D printing, enabled by automation, has the potential of allowing micro-factories and giving everyone the chance to become a one-person factory. What kind of new jobs will open up when anyone for a few thousand dollars can set up a little shop to make and sell the inventions they come up with without needing millions of dollars in capital to get started? Look at the size of the ‘maker’ movement to see where this could possibly go.
All those ‘green jobs’ many believe are coming? If they happen at all, it will be because factory automation has brought down the price of alternative energy products to the point where it’s feasible to eliminate old-tech power generation with high-tech computerized grids that can accept feed-in power from intermittent sources. Solar panels and wind turbines can only be competitive with coal and oil if made in high-tech, automated plants.
I could go on all day. It should be a trivial exercise for you to find examples of jobs that have not been hurt by machines, because after all, we only have 4.1% unemployment. The vast majority of people who want to work have jobs, even though doomsayers have been declaring the end of labor ever since the loom was invented.
I have shared several podcast episodes on this thread. Now I’ve just heard one that is IMO the best yet, particularly in terms of being so well aligned with the fundamental question: when this happens, will it be an economic apocalypse, or the beginning of a utopian era? This question is directly explored at the end of the episode. (It’s either going to be Burning Man for everyone all the time, or more like Third World countries where the superrich live behind walls and use helicopters to avoid the masses.)
The guest, Andy McAfee, is not someone I had heard of but it seems like I should have. He is the co-director of the MIT Initiative on the Digital Economy, and the associate director of the Center for Digital Business at the MIT Sloan School of Management, having formerly been a professor at Harvard Business School. Quite a resume, and his insights demonstrate why it is deserved.
I highly recommend listening to the whole thing, but some takeaways I thought were particularly notable:
–The people coding machine learning AI are using the same algorithms they’ve had for decades. The reason stuff like voice recognition is getting so much better is not the code, but the speed and power of the hardware (thanks to Moore’s Law).
–He talks about Moore’s Law in relation to the parable about the chess board and the grains of rice. He points out that on the first half of a chess board, you’d get up to a very large amount of rice, but it would still be an amount you could harvest from one field of a few acres in one season. It’s the second half of the board where things get out of control. And he believes in terms of AI, we’ve crossed to the “second half of the board” since roughly 2006.
–As an illustration of how this exponential growth has changed things so quickly, not just since the Apollo years that people normally cite, but much more recently: In 1997, some company (forgot the name) built a $55 million supercomputer that took up almost as much space as a tennis court. It could process some previously unheard of amount of data per second. By 2007, a decade later, the same processing power was found in any single XBox game console. Whoa.
–Therefore, due to the exponential increases in computing hardware power (even if Moore’s Law slows down a bit) and the fact that machine learning allows these AIs to teach themselves, he thinks we’re going to go a lot further a lot faster than most people think, and we’ll be seeing “stuff that looks like science fiction” on the regular.
Crikey, I’ve sure been! “You must complete 120 assignments per day.” No matter that some assignments are trivial and you can just cut-and-paste 'em, while others require phoning someone’s call center and waiting twenty minutes on hold. It was extremely procrustian.
Hellestal, I have been reading your posts carefully, with interest and–I believe–perspicacity. As I understand it, you and Sam Stone are making a similar argument, which I would summarize as follows (please tell me if you think this is inaccurate):
“AIs/robots/automation are not a serious threat to human employment unless and until we have cheap, mass-produced robots that are equal to or better than humans in every respect, including dexterity and adaptability to new tasks and challenges.”
I would agree that a company or business sector cannot be entirely automated so long as this is true. But if you can get rid of 95% of the humans (formerly) needed in fairly short order, the effect is almost the same. So you may have a few people overseeing the whole process, tending to the robots, making sure everything is running smoothly and intervening when they hit wrinkles they’re not prepared to deal with. But that may not be enough employment to have something for everyone to do.
Often I think the automation creeps up without people really noticing. A great example is shown by something Sam Stone said while trying to make a similar argument to yours:
I recently visited a relative in the hospital. He was referred there by an urgent care clinic due to an extremely high blood pressure reading. The hour or so I was there, the nurse only came in once. But the automated blood pressure device attached to him took readings several times. Presumably this was all being monitored by the nurse at the nurse station, who was able to cover many more patients than in the days when nurses would have to go around taking readings manually.
I wonder BTW why you feel the need to make comments like:
Sneering at other commenters in the thread collectively may not technically violate any rules, but it’s rude and unnecessary. And what’s wrong with a 15 (actually, 19) page thread about such an interesting and fundamental topic that updates itself periodically?
This was touched on in the podcast. Robots will be able to build bridges from scratch long before they will be able to repair those that are extant. So the MIT guy, who’s not a big fan of basic income programs, favors instead putting money into lots of infrastructure repair (similar to what someone suggested not far upthread).
Yeah, my impression had been that this kind of thing–excess “fat” in middle management, for instance–may have been prevalent in the mid-20th century in the U.S., but mostly disappeared once globalism became a thing and corporations got really cutthroat about hiring management/efficiency consultants to cut their workforces to the bone.
This looks to me like a strawman. I don’t think anyone in this thread is a latter-day Luddite, who wants to smash the automated machines that indeed magnify human labor and have great potential to raise standards of living for everyone (if governments adopt the right policies). If they are, I missed it.
But it should be noted that, as NPR’s Planet Money noted when they covered this topic and talked about the historical Luddites, the Luddites were actually not as wrongheaded as they are made out to be (emphases mine):
I don’t want to stand in the way of progress in AIs and automation that could make for a higher standard of living for everyone. But I would like us to think about how to make the transition smoother for 21st century unemployed workers than the British did for theirs in the 19th.
If I may, I would paraphrase it a bit differently. Let me explain it like this:
As humans we tend to value skills which only a subset of humans have. So, you can speak one language fluently (i.e. monolingual)? Big whoop, virtually everyone can do that. You can play complex songs on a piano? Wow, impressive (and potentially something you can do for a living)
But because of this bias we repeatedly make the wrong assumptions about AI. It’s much easier to write a program to solve a specific problem than it is to engineer a general intelligence with human-level cognition. So in terms of what’s “easy” and “hard” the tasks are largely flipped and a lot of the things *all *humans can do, turn out to be very hard to make a machine do.
We can make a burger-flipping machine. But can we make a burger-flipping machine that we can also give instructions like “Hey, can you give Bob a hand with a heavy box out back?” or just notice a spillage near the counter and it’s own initiative go clean it up?
Big IF.
I think you are underestimating how many jobs include that degree of flexibility I’m alluding here.
But better to say: AIs/robots/automation are not a serious threat to human employment unless and until they can purposed/repurposed to new tasks as easily as a human being can be so purposed. And the only possible way that they be so repurposed/purposed to any given human task is if (1) they have full human dexterity and adaptability to new tasks and challenges, or else if (2) the costs and timing of a new development cycle is about one trillionth what it is in reality.
You don’t need full-human dexterity if you build it from scratch. But what people don’t seem to realize is that building from scratch is a multi-year, multi-million (often multi-billion) dollar operation. That is not going to change. It’s going to get worse. The easiest stuff to automate? That stuff has already been automated. The low-hanging fruit has already been picked. Everything from here is going to take longer (on average) and cost more (on average) than the things that have already been automated.
More automation is going to happen, of course. Is happening. Right now. The next fruit higher on the branch is there for the taking, and somebody is going to pick it. But it’s never fast, never cheap, never easy. It’s not going to get easier. The easy stuff is already gone. It’s going to get harder.
And if you wave a magic wand that can cure all cancer, then all cancer would be gone.
There is NO POSSIBLE WAY to get rid of 95% of humans “in fairly short order” (absent strong AI). There is no industry that employs 95% of humans. There is no single piece of regular “automation” that can possibly replace 5% of humans across all industries “in fairly short order”, let alone 95%. Human work has specialized and divided into myriads of very, very different tasks, and those tasks cannot be automated simultaneously. It’s already getting into absurdity at 5%.
It takes too long. It is too expensive. Humans do too many different types of work. This kind of multi-purpose tech does not exist, and cannot possibly exist, because human work is not concentrated in any such way, and because automation of any specific task is an extremely high-cost, high-research affair. As has already been pointed out repeatedly. The existence of a “strong” AI is the only possible exception. Such a thing might, in fact, come into existence, as also mentioned repeatedly. But “jobs” are the least of our concerns if such a thing were to be invented. People worrying about jobs in a world of strong AI are putting on sunscreen to protect themselves from a nuclear blast. They are not interpreting the scope of that change correctly.
Outside that extreme case: new methods of automation are not developed so quickly that it can replace 5% of human work in one go, let alone 95%. Pure fantasy. Take, again, the transportation industry. Years and years of research, and billions upon billions of dollars, into automated transportation, and it doesn’t affect even 3% of the workforce, and even when it does, it can’t possibly roll out nation-wide simultaneously. The squeeze will be on those drivers, but it won’t be anything close to “in short order” change across the whole nation. It’s been years and years in the making already, and there will be other kinks to work out when it’s finally being implemented commercially. More than that, nobody can copy&paste the code for automated transportation in order to replace heart surgeons. The code will have to be written from scratch, and it’ll take another >15 years of research and more billions upon billions of dollars of investment. There is no one set of instructions that can do more than a small fraction of total human jobs, outside of a “general” intelligence.
We keep making this point. We keep making it because it is true.
Does the world look perfect?
Do you look around you and think, “Hmmm, things just can’t possibly get better for any human being anywhere on earth. We’re at peak civilization right here.” I’m thinking probably not.
These are equivalent arguments. “Not enough employment” is equivalent to saying “things will be so perfect, there will be literally no room for anyone to do anything else”. The entire reason jobs exist in the first place is that people are not satisfied with the world as it is now. As long as people look around and think that things aren’t perfectly swell, then there is a room for jobs.
That’s what jobs are. Literally.
Employment in the health care industry used to be around 2% of total employment in the 1950s.
Now it’s close to 10% of total employment.
You have recently visited a hospital, and on your visit you didn’t happen to see many nurses in that one context. It doesn’t seem to have occurred to you to wonder whether your experience was representative of nursing in the industry as a whole. Important fact: it was not. There are more nurses than there have ever been. I wouldn’t quite have written it the same way Sam Stone did. I would not have said there was “very little automation” in health care. Hospitals and health centers have machines out the fuckin wazoo. And also? The industry has more workers both in absolute numbers and as a percentage of total employment than they have ever had before.
The world is not perfect. We have jobs to deal with these imperfections, as we work for each other to make things a little bit better than before.
A big part of the imperfection of our existence is our frail human bodies. There is a reason why health care spending is such a large part of the economy, both spending on machines and also spending on health care employment. There is a reason why health care employment has become so large, both in absolute terms and also relative to the total workforce, even given the equally true fact that we use more medical machines than ever before. The frailty of our bodies is one of the major imperfections that we face. Jobs (and machines, too) exist to engage the imperfections of reality. Human health is one of the imperfections that is hardest to fix, and so we are throwing more and more resources – most especially the valuable resource of human work – at that major imperfection.
I make comments like this because they are true.
The belief that some piece of machinery might come into existence that replaces “95%” of human jobs “in short order” is exactly the kind of thing I’m talking about. This comes from the imagination, not from a sober look at the actual costs and timelines of automation as it works in the world (absent strong AI, which is exactly what is not being distinguished). Your previous post was about a podcast relating to “machine learning”, and I suppose the belief that a general intelligence might come along more quickly than anticipated. (I’m not inclined to believe that, but on that point, I could be wrong.) But “machine learning” is not the same topic as “jobs”. They are related, but they are still very different topics. The topics get conflated with each other, as I noted. And just as you have done.
It’s not a sneer to point out tendencies that have actually happened in this thread.
It’s most especially not a sneer when you give an example that specifically illustrates that very point. You provided that 95% comment on your own. You complain about my description of the thread, simultaneously as you provide more evidence that my description is accurate.
(SDMB default: 50 posts/page. Open a browser you don’t use, where you’re not signed in, and you can confirm.)
In general, there are two basic ways an extended discussion can go.
Poster A: “I think this viewpoint is reasonable.”
Poster B: “Well, here’s some facts that make me believe it’s not so reasonable.”
Poster A: “Oh, I didn’t know that. I’ll think about it and incorporate this info into my future posts.”
Alternatively,
Poster A: “I think this viewpoint is reasonable.”
Poster B: “Well, here’s some facts that make me believe it’s not so reasonable.”
Poster A: “Oh, well, in a few weeks I’m going to repeat what I previously said and ignore those facts as if I’d never read them.”
The first is better than the second.
It’s not a crime not to know something. Problems begin when facts are provided, arguments are made, yet the subsequent “response” does not incorporate any of the facts or engage any of the arguments. It’s always a better discussion when people inform themselves. For a specific example: I previously misread a post on “Labor always loses” to mean something pathologically non-factual. It turns out that I misread that post. So now? I’m not going to interpret that phrase from the same poster in the same wrong way again. (The phrase actually conveys an entirely different pathologically non-factual meaning. But I need to acknowledge that the deep wrongness that I previously inferred is not identical to the deep wrongness that was originally intended.) In your case? In future, I would recommend that you look up actual job numbers when talking about the automation of any particular industry like nursing, rather than relying on your personal experience of a single context. Your life is not a representative sample. While it’s true that we have many machines helping us with health care, it’s also true that we have more human beings than ever before helping us with health care.
Machines and people, working together. There are too many imperfections in this world to do anything else.
This is just make-work bias.
Does this guy think the world will be perfect? No?
Then there will be jobs. If we don’t have literally everything we want, then we will adapt to produce the things we don’t yet have. That’s what jobs are. The idea that we have to prioritize “repairs” is exactly the same imagination-based thinking. I mean, it’s possible that we should be prioritizing more infrastructure repair anyway, but the reason for that is not “jobs”. The reason is that the world is imperfect, and shitty bridges is potentially high on the list of imperfections. Infrastructure repair could be at the top of the list to move things in a slightly better direction, given our imperfect world. But if it’s not actually at the top of the list? Then we lose out by prioritizing it over other, more important things.
And here is a very important point:
As long as the world is imperfect, we lose out by using more workers on any given task than what is required for the task. This is something that is deeply non-intuitive, but still true. The world is imperfect in many, many ways. We work in order to remedy some of those imperfections. That’s what jobs are. But our pool of workers is limited, and can only remedy a limited number of the world’s imperfections. So if we make up worthless jobs that aren’t actually needed, or use more workers for one particular imperfection than is required to remedy that imperfection, then the available remaining workforce is much smaller, and therefore much less capable of remedying the many, many, many other remaining imperfections around us. We should use the minimum number of people possible to do any kind of work, like repairing our bridges, so that we can have a larger pool of remaining workers for the next most glaring imperfections on the list. This is true even when bridge repair (or anything else) is a high priority.
The attitude of “Let’s put people to work doing something unnecessary!” is deeply unhealthy, when so much necessary work remains around us, when the world is so imperfect. And in the majority of cases (tho not all) we should leave it to people themselves to decide what they feel is necessary for their own lives.
This is what is so amazingly faulty with so much of this discussion: the belief that there will be no more room for jobs is equivalent to the belief that we will build a perfect world. The belief that there will be no more room for jobs soon is the belief that we will build a perfect world soon. (Or that we will go extinct soon. That’s… the other possibility.) Jobs exist because we want more things (meaning: we exist in an imperfect world that falls short of what we want), and so we work to get what we want: we specialize in certain tasks in order to remedy at least some of the imperfections that are around us in our lives. If there is no more room for any jobs, that means that almost all imperfections have been solved. (There is a slight complicating factor of unnecessary job loss due to business cycle fluctuations. But that’s the result of bad monetary policy, not technology changes. Money, not automation.)
What I am NOT saying: all of this should be totally obvious on first blush to every person interested in this topic. Much of this is counter-intuitive. Much of this is built on evidence that is not obvious.
What I actually AM saying: when a plain factual point is made, it needs to be acknowledged for a discussion to be meaningful. The total number of jobs (increasing) and the wages paid (highest in history, globally) are facts. It’s similarly a fact that global wages are the highest in history because automation is the most extensive in history. Rich countries are countries with the most machines per capita. Simply a fact. The extensive costs and long development cycles of automation are facts, too. Many jobs can be automated, but fewer can be automated economically. None can be automated quickly. Which is another way of saying: it is difficult to become a rich country, because becoming a rich country means creating an incentive-environment where it is sensible to build machines, and building machines is hard. Human work is valuable, and it’s hard to replace. But even when it is replaced in one task, human work can be repurposed more easily and more efficiently than any single machine.
I don’t know of any quick way to dump a large load of weird, non-intuitive facts into a discussion. But I do know that a deeper discussion is difficult, if not impossible, when the most fundamental facts are rejected repeatedly.
That’s an interesting perspective on work, and in a better world I would agree. For most people, “not satisfied” means “not wanting to starve in the streets.” 36% of people in the U.S. don’t own their homes, and most of the 64% who do “own” them in the sense of having a 30-year mortgage and property taxes that require them to be working; or they’re homeless.
I agree, but I don’t see us shifting to a form of socialism any time soon; in fact, I think it’ll be quite a while after many millions of lives are ruined by not being able to provide for themselves. “Unnecessary” work is still preferable to no work. At some point after automation has taken over the bulk of human labor, I think we’ll actually be far better off. But the transition is going to involve some serious growing pains as a civilization as we alter our attitudes from “anything that happens to you from not having enough money is your own fault, you lazy bastard” to an attitude more in line with your own; that labor is something that should be applied towards the betterment of humanity and everyone is free to do so, because their basic needs are met by the robots and the Star Trek replicators breaking trash down to their molecular components and then assembling anything we need from them.
But this thread, as has been mentioned many times upthread, is about the potential realities of the transitional period between here and there. I think you’re a bit too concerned about the *when - it could be in 5 years, it could be in 50. I’d be shocked if it’s more than 100.
This is not at all what’s being discussed. It’s perfectly reasonable to assume a world in which very little human labor is needed anymore - but lots of *work still needs to be done, and is done, by fleets of corporate-owned robots who exist solely for the purpose of making money for the benefit of a very limited number of stockholders, and provide little benefit for anyone else.
What’s intuitive is the R&D model you’re describing; spend ten years and billions making a limited-purpose program, apply it to a specific task, spend another ten years making another limited-purpose program, apply it to a different industry, rinse and repeat.
That’s extremely linear. What’s counter-intuitive is exponential growth - Moore’s Law, as described by Slacker above with the example of a supercomputer in 1997 taking up a tennis court, versus the same processing power in a home gaming console ten years later. No one can say exactly when general-purpose A.I. will come along, because exponential growth is very counter-intuitive. Someone could make the breakthrough tomorrow, because what was missing was hardware powerful enough for their software to make that leap - or it might be 50 years from now. We just don’t know.
And that trend will continue, until the point where automation becomes so cheap and widespread, and the applications so diverse, that it simply makes no sense to use a human for most work. And that day may be a lot sooner than most people think, because of Moore’s Law.
Hellestal, I’m going to try to keep this brief and not respond point by point, because frankly as I intimated upthread I’m not keen to carry on an extended conversation with someone whose debating “style” consists of “I don’t care if they’re from Harvard, MIT, or Princeton or have won Nobel Prizes; anyone who disagrees with me is stubbornly unwilling or sadly unable to comprehend the basic facts I’m laying out.”
It seems increasingly clear (although you have not quite said so explicitly, I’m reading between the lines) that you are an advocate of a brand of extreme libertarian, laissez-faire economics that utterly rejects Keynesianism. You may even be an Objectivist (i.e., a follower of Ayn Rand). That’s fine: everyone’s entitled to their own ideological viewpoints. But your insistence that these are just the plain and obvious facts of the world, and refusal to even implicitly acknowledge that many of your views are actually out of the mainstream, is irksome and not an attitude, as I say, that I care to engage with. I have some non-mainstream views myself, but I hopefully don’t go around sneering or sighing at people as though it’s a travesty that I should even have to explain or argue for viewpoints that are so plainly True.
ETA: AI Proofreader, great post just above mine, except that I think you give too much ground in letting what’s proposed be dismissed as makework. Repairing bridges across the country would IMO be anything but! A lot of what was done in the New Deal jobs programs could be seen as makework, but we continue to benefit from the rest areas and other nonessential infrastructure they built.
However, considering the larger thrust of that post I was willing to concede that there could easily be considered more worthy projects to devote ourselves to. Especially from a free-market capitalist perspective; there’s not much money to made in it, and it’s not a novel or innovative thing necessarily. It just needs to get done.
However, considering the larger thrust of that post I was willing to concede that there could easily be considered more worthy projects to devote ourselves to. Especially from a free-market capitalist perspective; there’s not much money to made in it, and it’s not a novel or innovative thing necessarily. It just needs to get done.
Yeah, my wife to whom I’ve been married now for nearly a decade arranged to meet IRL for the first time in Minneapolis/St. Paul on the day of the big bridge collapse. Fortunately neither of us was crossing it at the time, but it was freaky—especially since at one time I lived right at the entrance of that bridge, so close that my old apartment building was blocked off by the police. So I definitely support a massive national program of bridge repair, irrespective of its utility as a jobs program.