Will robots take away all our jobs? Steven Rattner says no:
Almost exactly 60 years ago, Life magazine warned that the advent of automation would make “jobs go scarce” — instead, employment boomed.
Now, the launch of ChatGPT and other generative A.I. platforms has unleashed a tsunami of hyperbolic fretting, this time about the fate of white-collar workers....A breathless press has already begun chronicling the first job losses.
....[But] when it comes to the economy, including jobs, the reassuring lessons of history (albeit with a few warning signals) are inescapable. At the moment, the problem is not that we have too much technology; it’s that we have too little.
....Higher worker productivity translates into higher wages and cheaper goods, which become more purchasing power, which stimulates more consumption, which induces more production, which creates new jobs. That, essentially, is how growth has always happened.
Rattner takes us through the usual well-worn history of the abacus and the Luddites and his old handheld calculator to make his point. Sure, these things put some people out of work, but eventually new and more plentiful jobs took their place. "This is how growth has always happened."
I've just about given up trying to understand how smart people can believe this. We've had precisely one (1) genuine Industrial Revolution, and yes, it turned out fine. But why does anybody think that n=1 is predictive of every future revolution?
The Industrial Revolution mostly replaced muscle power. We still needed lots of brains to keep running things. The Digital Revolution will replace brains too. By definition, if we eventually develop true artificial intelligence it will be able to do anything a human brain can do.¹ This means, by definition again, that any new jobs created by AI can also be performed by AI better and more cheaply than people. There will be virtually nothing left for us.
How is this not obvious? Help me out here. We could, of course, address the AI revolution by flatly banning robotic labor in a wide range of situations. Or maybe we'll end up with worker riots, except putting the torch to robots instead of powered looms. Who knows?
But short of artificial solutions like that, true AI really has no plausible endpoint except mass unemployment. This will obviously require a huge rethink of how we compensate people once hourly work is out of the picture, but hourly work will indeed be out of the picture. People really need to pull their heads out of the sand on this.
¹If you don't believe true AI is possible, that's a whole different argument. It's probably not one you can win, but at least it's an argument.
This is worth a listen:
https://www.econtalk.org/marc-andreessen-on-why-ai-will-save-the-world/
The only reason to listen to that grifter is to find out what he's invested in this month. I'm serious - go back and look at what he was pumping and dumping this time last year.
Also, Kevin is a bit too blithe about his:
The thing here is, everyone ignores the transition. Sure, it is fine now, but we had kids pushing coal carts for around 120 years during that "transition", which lasted several generations.
It "turned out fine" if you lived after it happened; if you were part of the excitement, you might feel a little different. So I don't know if our retiree is affecting the voice of whatever comes after all of us or what, but if you're alive right now, you might worry a bit more about the "transition" than your post-scarcity grandkids will.
Right! It's the classic "the ocean will be calm after the storm has passed". Why, thank you for that.
The Industrial Revolution wiped out a class of workers starting in the late 18th century, and it wasn't until 1850 or even later that it started raising living standards from the new minimum. Good luck if you are out of work for 70 years. Maybe you can keep your eyes on some distant prize in the far future like a 19th century capitalist or Stalin, but not everyone can afford to do so.
Kids often worked terrible, dangerous jobs before the Industrial Revolution. Many of them died young -- very young. Children in medieval and early modern societies often led absolutely terrible lives.
The Industrial Revolution did not immediately make that better, and it arguably made work life worse for the very poor, but it was accompanied by greater life expectancy and greater opportunity.
IIRC, there was a decrease in children's heights during this time.
"We've had precisely one (1) genuine Industrial Revolution, and yes, it turned out fine."
I am not sure it was fine. Certainly, expansive life-saving and life-changing technologies, but also the joys of industrial mining, industrial waste, industrial accidents, mass consumption, and a changing climate. We also transformed the planet, ourselves, and our societies with the Neolithic Agricultural Revolution. There is little doubt this new (third) transformation will be even more transformative. AI is certainly a part of it. All life tends to do this. (see the Great Oxidation Event). The trick is not to destroy yourself in the process.
Ok. No more Luddites. How about “Coners?” https://sfist.com/2023/07/07/pranksters-claim-they-can-stop-self-driving-cars-with-orange-cones-waymo-and-cruise-not-amused/
My knee-jerk reaction is to agree, but my contrarian gene kicks in.
What happens to the people whose comparative advantage is muscle power? The new jobs likely will require less muscle and more brain than the old jobs. Likely there will be two types of losers from innovation like AI--one is those whose old jobs are eliminated and are too old or too inflexible to qualify for the new jobs. Time will cure that problem, though the pain will be real. The other is those who can't qualify for the new jobs, ever. Those must depend on the social safety net.
The thing is for a while its the smart people who are out of work. Say you're a janitor. Yoy need fine motor movements, navigate areas whose layout may change daily, stairs and so on. It is far FAR cheaper to pay someone 35k a year than to make a robot that can do all that.
It is far FAR cheaper to pay someone 35k a year than to make a robot that can do all that.
Yeah. I think this is where maybe Kevin's vision falls short. It's true that machines are far better than humans at plenty of things, but in many cases "far better" isn't necessary and "merely ok" is perfectly acceptable. In those cases where it's the latter—and the human is simply flat out cheaper—human workers may enjoy lasting staying power. Also, even the industrial revolution—which as Kevin rightly points out was all about doing physical activities faster and more powerfully than humans and animals—the fact is millions of workers even in rich, capital-intensive countries like the US still focus mainly or considerably on physical tasks. So, if the first industrial revolution didn't obviate the needs for human muscles, why are we so sure the AI revolution will obviate the need for human brains?
Over a ridiculously long timescale.
The Industrial Revolution threw, conservatively, six or seven generations worth of people into its gaping maw, chewing them up and devouring them, before its benefit started to be anything like evenly distributed. If you were a modal British person in the 18th and 19th century, the Industrial Revolution probably LOWERED your health, your life expectancy, your standard of living, and your control over your own life.
A Manchester mill worker in 1800 could look forward to his son, his grandson, and his great-grandson being worked into early graves in dangerous circumstances before things would meaningfully even start to change.
Even in seems to be counting every technological revolution from James Watt down to the Internet revolution or the 90s as one event.
And it only started to get better after the violent Chartist movement that no one likes to talk about, perhaps because it was violent and had a measure of success or perhaps because a feudal leader, the King, stepped in and threatened to pack the House or Lords if working conditions weren't improved.
It turns out we don't need jobs anyway. Not in my backyard.
https://www.reuters.com/markets/us/new-twist-nimby-no-factory-my-backyard-2023-07-10/
MARSHALL, Michigan, July 10 (Reuters) - Fred Chapman has a message for Ford Motor Co (F.N), which is planning to build a sprawling factory on the outskirts of this town to make batteries for electric cars and which promises to employ 2,500 people. “We don’t need jobs,” he says.
Hilarious.
With AI, the short term disruption for people working today will be like the China shock and they will never recover. As long as we're OK with that.
The Industrial Revolution mostly replaced muscle power. We still needed lots of brains to keep running things.
Yeah, unfortunately, as it turns out, a good 50-75% of people have more muscle power than they do brain power. (This isn't to say that lots of people are "strong" or that lots of people are "stupid." It's to say that, for a lot of people, they're better at doing physical tasks - carpentry, car repair, hairstyling, stocking shelves, baking, whatever - than they are at doing mental tasks all day long.) And we still haven't fixed that employment mismatch yet - except for the brief period we're in right now, generally for the last century there have been far more people looking for physical labor jobs vs. mental labor jobs.
I'm nowhere near as bullish on AI working in the future... probably because my job is computer-based and requires a lot of input from a human being to get the so-called "automated" parts of my job to work correctly. But, sure, if you have a room full of monkeys typing every minute of every day for decades, eventually one of them will perfect AI code.
So let's assume Kevin is correct that eventually AI will take over lots of mental labor jobs. That still leaves creative labor jobs - I don't know that AI is going to be able to optimize what today's latest clothing fashions should be, or write all the music that kids want to listen to or whatever, since a lot of that is highly celebrity-specific, in that customers want the product because a highly-influential person says it's cool. It also leaves service labor jobs - I don't know that robots are going to be able to do the whole "caring for young/elderly/sick people" thing very well. I mean, we had a test run with the pandemic, and everybody apparently hated not being able to see a real person face-to-face for their medical and educational needs. (Yes, I understand that Japan gave lots of animated stuffed animals to comfort its old people and they love them, but I still don't think that that's going to be The Solution to taking care of our elderly here.) And finally, it leaves entertainment jobs, which generally people want to see other people doing. (Watching cartoons is OK, but I doubt that many people are going to want to watch robots or CGI versions of actors doing stuff for all their TV/sports/movie/TikTok watching.)
I'm pretty sure with the transition away from doing physical and mental labor stuff, we'll get stupider and more helpless as a population, but I doubt that jobs are going to completely vanish. With everybody home all day, there's going to be a lot more demand - even more so than today - for services and entertainment.
oh, you beat me to it! Pretty much in accord with this.
Re: monkeys banging on keyboards - Isn't the rule of thumb that a million monkeys banging on keyboards will produce a LINUX kernel every 23 days?
Eh, I definitely prefer CGI to real people visually but folks still need to be in the suit, VAs MoCap. Code Miko is a good example.
Kevin isnt really making an argument or a coherent case for why this will happen.
He is simply begging then question and cant understand why others dont simply assume the conclusions that he prefers.
Assume we create a tool that can do everything better than a person could, then there will be nothing that people are better at! If we assume that AI will be better at every job, then people wont be better at any jobs! If you assume mass unemployment, the result is mass unemployment! HOW CAN YOU NOT GET THIS!!!
Obviously everyone doesnt share this basic assumption. Yelling at them a bit louder and insisting that they accept the assumptions probably isnt going to change any minds or cause anyone to rethink anything.
Kevin's posts about AI have always been dumb. In fact they are so dumb the make me wonder if his posts on other topics are equally dumb and I am just not smart enough to see that.
Are you saying that self-driving cars aren't all over our roads right this second? Because Kevin assured us they would be by now.
Please God, tell Kevin to come off his high horse and quit thinking he can predict the future. Or if God is actually giving Kevin the information about the future, let Him tell the rest of us that He is doing this, and that we must follow Kevin like Moses.
Kevin says AI is something completely new and couldn't be the same as what happened in the past. OK, if it's completely new then how is Kevin able to predict how things will turn out? This may be prophecy, but it ain't science. Science uses reproducible phenomena to formulate "laws" so that future phenomena can be predicted and controlled. The laws aren't formulated out of thin air, they use observations from the past. Historians try to do something similar - they try to identify patterns in the past and get some insight into how things will go in the future.
People have been predicting the advent of "robots" for a long time, but the supposed loss of employment would be a result of greatly increased productivity and instead of increasing productivity has been decreasing. There is no evidence as yet that AI of any kind has or will increase productivity. And again, despite the huge gains in productivity from mechanization all through the industrial revolution there has never been an increase in unemployment.
Kevin is right that we need a "rethink of how we compensate people". For a long time real wages actually kept up with productivity but this changed suddenly fifty years ago. Also work hours decreased until 1940. If "robots" do actually increase productivity then there is no real economic reason that wages per hour shouldn't increase and hours worked decrease - these things actually happened up through the middle of the 20th century. Things changed at that time largely for political reasons in the US as the collective power of wage-earners was decreased, although international competition also had something to do with it. To shift the balance back toward wage-earners may require winning or avoiding the culture wars.
What actually happened after the Civil Right Era was that the political power of the wage-earning majority was split by the deliberate policy of Republicans to support the continuation of White Christian Superiority. Getting past this political development, which so far has culminated in the election of Trump, is probably more important that any disruptions from AI.
I also think that white Xtian nationalism is the number one threat facing the humanity today. AI takes a distant not-first, and is probably not even in the top three. The political consequences of that terroristic movement and modern capitalism are far-reaching and immense. I can't think of a single issue (including AI!) where those two things don't influence the outcomes at a fundamental level.
Unfortunately there is tribalism/nationalism/racism everywhere. This is frequently combined with religion sectarianism - it's basically all the same set of group cohesion instincts that all social animals have.
And there are leaders everywhere, like Republicans, who cynically exploit this for their own advancement, although some really believe they are chosen somehow to lead their people to victory over the "others". Keeping this contained is a constant struggle.
Mote than climate change? We need a planet to fight over
"By definition, if we eventually develop true artificial intelligence it will be able to do anything a human brain can do"
I generally agree with you, but not with this definition of "true" artificial intelligence. It seems very possible we could create very powerful AI, even sentient/self-aware AI that is just wildly different than human intelligence and not good as some thing we are good at.
It seems very possible we could create very powerful AI, even sentient/self-aware AI that is just wildly different than human intelligence and not good as some thing we are good at.
Why would this be, though?
Unless you're talking about a ghost in the machine, there's no reason to think of the human mind as anything but a manifestation of the human brain, and the brain is simply a very complex machine (though one composed of biological materials and not silicon).
In other words, given enough time, why couldn't a machine be produced that literally can do anything a human brain can do?
In theory I don't see why literally anthing a human brain is capable of is not replicatable (given enough time).
I have no idea why nor do i particularly expect it to be the case, but from the little I know of current AI research and development no one really knows, because we don't understand very well how the human mind works or even how some of our current AI systems really work. So while in theory i also don't see any reason it isn't possible to create an AI to replicate all human intellectual capabilities I'm not ruling out that we may fail at it. Maybe it's just beyond our technical capabilities and will remain so, even if it's theooretically possible. Or maybe it will just prove too expensive to be worthwhile.
The brain is not a machine. It is an organ.
what do you think the difference is between a machine and an organ? why is an organ anything more than a biological machine?
"By definition, if we eventually develop true artificial intelligence it will be able to do anything a human brain can do"
Does that include spreading kuru?
One thing that won't happen is that there won't be believable androids for quite a while so in any circumstance where you enjoy interacting with live humans jobs might expand.
Live theater might have a new popularity, live music, writers who can be very spontaneous might find audiences who begin to enjoy their interaction.
Machines were invented to be stronger and faster than humans but that hasn't stopped anyone from working out or playing sports.
AI generated trailer for Heidi,
https://twitter.com/karpi/status/1678372061176004609
It's the uncanny Alpine valley. Beat that, humanity!
I would think that until you get true AI, there will be some uncertainty over exactly what AI can take over and what it can't. I don't know how that will turn out -- I would guess a small number of skills will be in high demand, and a great many more won't.
When you get to true AI, then don't you run into issues about sentience? Wouldn't those AI have rights? Unions? Minimum wage? I think if an AI can do everything a human brain can, that AI would deserve certain rights. I think what the world will be like when we get that happens is too hard to predict, but I also think it's further away than you do.
“By definition, if we eventually develop true artificial intelligence it will be able to do anything a human brain can do.¹”
So: a few things. I remain uncertain what the word “true” is supposed to indicate in that sentence.
Another important word in that statement is “artificial” (and, yes, it’s humorous to see “true” applied to something “artificial”). If the intelligence really is “artificial,” no matter how true it may be (whatever that means), that would seem to be an important qualifier. Because ultimately it remains difficult to see how an artificial version of something will replace the real thing in all of its manifestations.
By definition, then, AI would seem to be limited. But what actually is that definition? I’m sure there are several, but here’s the wiki version:
“Artificial intelligence (AI) is intelligence demonstrated by computers, as opposed to human or animal intelligence.”
So, there are three types of intelligence given in this definition. And we already know that the human and animal intelligences are quite distinct and cannot fully replicate one another. Why would we assume that a third type of intelligence actually can, in fact, entirely encompass one if not both other types of intelligence?
It seems more plausible to me to assert that AI, like other types of intelligence, will have its own set of pros and cons. How exactly those variables will play out in practice: I don’t know. But I remain highly skeptical that all humans will simply be left with nothing to do. In another hundred years (after AI has figured out how to keep me alive that long) I fully expect to be typing out these comments on my own. Why would I want it any other way?
So called artificial intelligence is certainly artificial but it is not intelligence--even the nifty stuff that the current generative AI stuff can do is not properly intelligent. I continue to say that "simulated intelligence" would be a much better term for what we are seeing now.
As for Kevin's assertions about "true" AI, I mean where to begin? First of all the proper term (and this is not one I am making up) is "General Artificial Intelligence", not "true" AI. Then he starts talking about robots, as if AI and robots are the same thing. Maybe he's just seen Terminator one too many times.
"How is this not obvious? Help me out here."
Glad to help. Please bear with me. And, yes, I recognize I may be overlooking something.
The fundamental Luddite argument is that things used to be great, now things aren't great, and we need to make them great again by turning back the clock. The logical flaw is that things change, we have to respond, and our only choice is to respond in a way that is adaptive or maladaptive. Trying to turn back the clock is the opposite of adaptive.
Name one non-human genetically heterogeneous social system. The reason you can't is because they don't exist. If animals take care of each other, it's because they're following a genetic instinct. The non-human social system instinct says, "Take care of family."
A troupe of chimpanzees is a fission-fusion society, not a social system. There is a chimpanzee social system, and it's a mother and her juvenile offspring. Other relationships within a fission-fusion society are purely transactional.
The origin of the hunter-gatherer band and the origin of our species are coincident and the effect of a common cause, which is the "wisdom" instinct that tells us to care about our family and our neighbors.
We are the only species that forms a genetically heterogeneous social system. There used to be many band-sized systems. Then, because they were so successful, and because the world is not infinitely large, they got larger and fewer in number. Now there is one system.
Taking care of each other is what our instinct is telling us to do. That's why it feels good. Ergo, there will always be lots for us to do, unless we keep doing what we're doing now, which is ironically something else that's so important that we don't have the time or the resources to take care of each other, not to mention that the "something else" is destroying our planet and tearing our social system apart.
The way I see it, this is what every legitimate religion is telling us. The Golden Rule, right? And isn't it what science is telling us to do? Don't prejudge? Don't jump to a conclusion and then go looking for evidence to confirm and defend your conclusion? Which might be that the Biden's are a crime family, or that Republicans are a different species. I think science is telling us that, if people disagree with each other, then look at the evidence first, and then draw an evidence-based conclusion.
I am afraid of AI, but not artificial intelligence. The AI I'm afraid of is artificial ignorance. Not being wise is being ignorant, and vice versa.
How is this not obvious? Help me out here.
As long as there is not a flood of low-cost, advanced humanoid robots, it seems most jobs today will still be in existence. Jobs will draw heavily on AI-based tools for help, and thus humans will get more productive and efficient.
I bet that 50 years from now, when the economy is not in recession, the unemployment rate will still be below 10%.
To me, the biggest concern is not an avalanche of job losses. The biggest concern is some Skynet-like scenario.
I can imagine it's technologically possible that one day a self-driving mower will arrive in my street on the back of a self-driving vehicle, get itself off the tray, and cut the grass on the public walkways, pausing when animals or pedestrians walk near. It might even be able to sense when a branch has come off a tree into its path and have a robotic arm to carry it back to the truck. I don't imagine it will be able to chat to local residents who want something done about the blocked gutter, but no doubt there'll be an app for that.
I do have trouble believing it will be more economical for local government to buy, operate and maintain a fleet of machines to do this, along with others that can lop trees and trim hedges and clean up fallen palm fronds and countless other tasks, than it is to employ the existing team of multi-skilled human workers who turn their hands to all sorts of outdoor jobs. The same applies to all sorts of service work; because a machine can do something doesn't mean it will be economically feasible to buy it. Designing and manufacturing and maintaining a robot that could do everything a hospital nurse does might one day be technologically possible, for example, but surely it would also be horrendously expensive. Ditto robots that could take over the work of a kitchen hand, let alone a chef; or a childcare worker; or a fitness instructor; or a border patrol agent.
Consequently I think the predicted job losses from AI-enabled automation will not be as great as Kevin seems to think. They will likely be concentrated in jobs which consist mainly of processing information, and may well have the beneficial effect of influencing people to switch careers to alleviate the chronic skills shortages in many industries. Many jobs will continue to be filled by humans for the simple reason they will be cheaper than any alternative.
Now instead of advanced lawnmowing robots, imagine a scenario with advanced robotic surgeons which are beyond humanly precise, have immediate access to the diagnostic knowledge of the latest medical research, and never tire and see what the cost-benefit analysis looks like.
They're banking on our pancreases. Machines replaced our muscles and bones in the Industrial Revolution. Computers will replace our brains and neurons, or at least drive us nutcase, in the AI Revolution. What's left? Why not the pancreas? It's a marvelous organ. Let's see them replace that with some new technological doodad.
AI costs a ton of energy to run. And currently a ton of people too. Chips will get a little more energy efficient--but we are running into physics.
Well, don't look now, but humans "cost a ton of energy to run," and there is a pretty infamous historical example that resulted from this recognition. I'm not sure we win out in this comparison.
Thanks, everyone, for pointing out that the transition can be very messy and last longer than a lifetime.
But also thank Kevin for ridiculing
"We've had precisely one (1) genuine Industrial Revolution, and yes, it turned out fine. But why does anybody think that n=1 is predictive of every future revolution?"
Let's see, seat belts have done precisely nothing for me since I've never been in a car accident. And n>>1. Since that's "turned out fine", maybe I'll stop using them.
Don’t get distracted by “true” AI or sentience. All we need is an AI capable of 75% or so of human ability. That will be good enough to permanently replace many, many millions of workers (probably billions). Thinking these people are all going to be baristas, service workers and actors is nuts. We will be faced with a massive unemployment problem, and the AI will only get better and probably very quickly. We need to start thinking about this seriously now. Universal Basic Income and some kind of tax on AI labour needs more attention.
Three thoughts:
Also by definition, anything a computing machine can do is Turing computable, which is to say that it has algorithms to complete, aka effective procedures, which is therefore limited to some extent. The limitations may ultimately be trivial, but we don't really know if human intelligence has additional capabilities.
We do not, at the moment, know all the details of how human intelligence works; we do have extensive knowledge of how certain abilities map, and we know some of the limitations of those capabilities, but the whole connectivity is still murky.
Is self awareness necessary for some things, and if so, is it possible to imagine that a machine can be self aware?
Note that there is a huge science fictional literature about the era of automation and a somewhat narrow literature about robots (Asimov limited the field enormously by his three laws, which require constant interpretation, resulting in many clever stories.
Working part-time, I bring in more than $13,400 every month. I made the decision to research it after hearing a lot of people talk about how much money they could make online. All of it was real, and it completely 10 altered my life. You can read this article for
additional information…. https://needpeopleNYC.blogspot.com
AI really has no plausible endpoint except mass leisure. AI will be needed to overcome the ruling power elite who only know how to enforce market discipline and devise an AI political economy.
Does this robotic system inspire you to imagine a wondrous future or does it fuel your nightmares?
I'll help you out.
The AI you're speaking about (here, at least; you kind of slide back and forth on your implicit definition of AI) is a lot harder to accomplish than you think it is. Machine learning is not even close to the end of the game as any machine learning expert will gladly tell you. You just have a habit of tuning out expert noise in this particular area. Bizarre. You, of all people, should understand that expertise matters.
The Luddites were protesting the effect of technology on their wages, not technology itself. The complaints about the effects of AI are frequently that.