[This is an entry to the 2019 Adversarial Collaboration Contest by Doug Summers-Stay and Erusian]
Adversarial collaboration on the question: “Automation/AI will not lead to a general, sustained economic crisis within our lifetimes or for the foreseeable future. Automation/AI’s effects into the future will have effects similar to technology’s effects in the past and, on the whole, follow the general trend.”
Defending the proposition: Erusian
Challenging the proposition: Doug Summers-Stay
tldr: Until the pace of automation increases faster than new jobs can be created, AI shouldn’t be expected to cause mass unemployment or anything like that. When AI can pick up a new job as quickly and cheaply as a person can, then the economy will break (but everything else will break too, because that would be the Singularity).
Introduction
As software and hardware grow more capable each year, many are concerned that automation of jobs will lead to some sort of economic crisis. This could take the form of permanent high levels of unemployment, wages that drop below subsistence levels for many workers, or an abrupt change to a different economic system in response to these conditions.
This has become a talking point outside of economic circles in the U.S. Democratic presidential candidate Andrew Yang’s most well-known policy proposal is a universal basic income to offset this (an idea Elon Musk has supported for years). Bill Gates suggested that when robots replace workers, the companies should be taxed at a similar rate to the taxes being paid by those workers. These entrepreneurs have spent a lot of time thinking about and planning for the future, and have a lot of experience with introducing new technology. Are their concerns valid?
Throughout this discussion, we use the words AI, automation, and robots more-or-less interchangeably. Imagining Asimov-style androids with positronic brains makes it easier to picture a world where all jobs are automated. In reality, though, it would be a silly waste of resources to literally have robots come in and do jobs as drop-in replacements for workers, and there are few jobs where this would make sense. A lot of software in the future will be more human-like in the sense that many machines could have natural language and image-understanding capabilities and have the ability to reason about the wider context in which their work exists to avoid dangerous or costly mistakes due to a lack of common sense. In many other ways, though, software for nearly all working robots will not be similar to human minds at all.
Some jobs, like picking most fruits and vegetables or most assembly line jobs, currently can’t be done by machines because they require manual dexterity. In such cases, you would need more precise manipulators with touch sensors that can adapt to a wide range of situations. There is no reason to expect they will look like human hands, though. Machines will also be networked, of course, so imagining them as a bunch of individuals is also unrealistic. Finally, the way tasks are currently divided into jobs makes sense for human workers, but wouldn’t make sense for automation. Instead, certain tasks will be automated first, and the remaining tasks that form part of a job will still be done by humans.
What is happening in technological unemployment today?
Worldwide, employment rates are not worse than historic levels. This seems to show that jobs are being created more-or-less as fast as they are being automated. Scott has covered this issue thoroughly in a survey article. The Bureau of Labor Statistics puts out a report of their projections for what jobs will be lost and what will be added for the next ten years. Some jobs that are expected to grow are in health care, renewable energy, and several computing professions. The ones that are declining include secretaries, clerks, assemblers, and care of outdated tech like locomotives and wristwatches. Automation has tended recently to take middle skill jobs. This has caused many people to take less desirable jobs.
We have seen at least three instances in the past where automation has taken a significant fraction of all existing jobs. Before the neolithic agricultural revolution, nearly everyone was a hunter or gatherer or both. During the period from about 8000-4000 B.C., this gradually shifted so that most people were employed in agriculture. Before 1400, around 70% of all employment was in agriculture. Today, it is only a few percent in advanced economies. At its peak during World War II, nearly 40% of U.S. employment was in manufacturing. Today, that number is below 10%. Housework also declined from 60 hours a week in 1900 to only 15 hours a week today (although this doesn’t show up in employment figures, of course, and can also be partly attributed to declining fertility rates.)
Another line of evidence to consider is GDP. Before 1000 AD, the per capita GDP everywhere was below $1000, adjusted for inflation. In Western countries today it is around $50000. Since human innate capability hasn’t changed, this must be the result of innovations (in education, processes, tools, machinery, or what-have-you) that allow people to produce more value for each hour of work. By one way of looking at it, this means that there are already 49 “robots” worth of automation for every person in these countries. Yet employment is still at roughly the same level it has always been.
So we know from experience that employment is able to adapt to extensive automation. However, these changes took place over millennia, centuries and decades respectively. If the pace of job automation were to increase so that replacement of a significant fraction of the workplace happened over years instead of decades there might not be time for people to retrain fast enough to avoid some higher rates of unemployment. Again, though, that doesn’t seem to be the case at the moment.
Are there things computers will never be good at?
A common response to the question of whether we will eventually reach the point that all jobs can be automated is to name a skill that computers will never be able to perform, attempting a disproof by counterexample. The following terms link to places a skill was claimed to be impossible for computers: write jokes, write novels, express compassion, robotically navigate a human environment (like an arbitrary kitchen), manufacture new categories at arbitrary levels of abstraction, act creatively, represent and invent concepts, learn from small data, express emotion, have motivational direction, think socially and cooperate.
Researchers are, however, aware of these limitations of current machines and actively trying to find ways to automate them. Here are the same terms along with a link to a paper where research is presented on how to automate the task: write jokes, write novels, express compassion, robotically navigate a human environment, manufacture new categories at arbitrary levels of abstraction, act creatively, represent and invent concepts, learn from small data, express emotion, have motivational direction, think socially and cooperate.
In every case, although researchers haven’t yet solved the problem and in some are far from a solution, it is possible to see a clear research direction and a path to gradual progress. Many claims of this sort are (in the local argot) making a motte-and-bailey argument. When someone argues that a machine can’t really feel emotion, they can always retreat to the motte that the machines are incapable of phenomenal conscious awareness of what it is like to feel (for example) emotional pain. This is true: we have no idea how to make a machine that is conscious in this sense or to test whether phenomenal consciousness is present in a person, animal, or machine. They then, however, make claims that the robots will not be able to respond to anger in a voice, or anticipate that taking an action might cause someone to feel sadness. This is false: a neural network trained on examples of anger in a voice could learn to discriminate it without the ability to feel blood rush to its ears. For the purpose of taking jobs, an accurate discriminator or ability to accurately simulate emotions is all that is necessary.
The new techniques researchers have come up with are not only able to perform better than previous methods on a particular benchmark; they are also becoming more general. Artificial General Intelligence is what human-like AI is usually called these days because the ability to handle unanticipated situations is such a central part of what makes human intelligence special.
The new Transformer neural net architectures are an example of how AI is becoming more general. Although simply trained on predicting the next word in a sequence, such models have demonstrated superior performance on question-answering, common-sense, categorization, and other benchmarks. A similar architecture, with few changes, can be used to compose music, create artwork, simulate voices, and so on. These models work well because they are able to learn to direct attention to the parts of the context most applicable to deciding what the next output should be. In the future, we should expect systems that are more adaptable still. An adaptable AI will be quicker and cheaper to deploy on new jobs, so we should expect the rate at which jobs are automated to increase.
In terms of hardware, we can expect computational capacity on the order of the human brain in supercomputers in the next five years and in home computers about twenty-five years after that. (Assuming 100 billion neurons, 1000 synaptic connections per neuron, 10 floating-point operations per interaction, and a temporal resolution of 1000 interactions per second.) So hardware shouldn’t be a limiting factor after 2050 or so, as long as current trends hold.
This is not to minimize how far we are at the moment from a machine that can learn an arbitrary new job as easily as a human can. People can model another human by “putting themselves in their shoes.” All of that ability to anticipate how other humans would react to an action has to be built into machines to achieve the kind of autonomy we are imagining. While machines can now act in creative ways as well as rational ways, tying the two together is still a very open problem. Systems that have lifelong learning, that continue to grow with experience, are still very rare. The ability to understand spoken or written language is still at a very primitive level. These problems don’t seem insurmountable– merely very hard.
What jobs will be automated?
Frey 2013 characterizes which jobs are likely to be automated soonest based on the following capabilities required to perform the job:
– Finger Dexterity, Manual Dexterity, Cramped Work Space, Awkward Positions
– Originality, Fine Arts
– Social Perceptiveness, Negotiation, Persuasion, Assisting and Caring for Others
He concludes that 49% of U.S. jobs are repetitive, don’t require fine dexterity, originality or people skills, and are therefore likely to be automated in the next few decades. This includes most office and administrative support jobs, sales jobs, some service jobs, and most production and transportation jobs.
However, Arntz et al argue that this number is much too high. Holding everything else the same, they show that this is neglecting the variation within a profession and the ability for a job to adapt when new technology becomes available. With these taken into account, only 9% of jobs are found to be at risk.
Both of these papers are discussing the right side of the graph above, everything above the “75% probability of computerisation” line. Eventually, though, essentially all the skills described on this chart will be automatable. While it may not fall within our lifetimes, it does seem to be part of the “foreseeable future.” It doesn’t seem like there are any fundamental physical limitations preventing it (in the sense that we may never build a spaceship that goes faster than light.) The existence of human brains shows that the right arrangement of atoms can compute at human levels with reasonable size, weight, and power restrictions. It seems reasonable to suppose that computers will continue to increase in capability until they will be able to perform any intellectual task required in a job as well as a human. This includes creative, decision-making, and emotional reasoning tasks.
To replace people in jobs also requires a body that can perform tasks with the dexterity and ability to adapt to different conditions that are required for a job. This also seems to be at least decades away for many jobs.
Beyond the invention of hardware and software capable of performing these tasks, the cost of developing and deploying the technology must fall below the cost of hiring workers in order for the workers to be replaced. The price of computing has been dropping steadily for several decades now, and there are no fundamental physical limitations to this improvement that would prevent the trend from continuing to the size and power-usage levels of a human brain. Robotic bodies and manipulators, while continuing to improve in dexterity, sensing ability, and cost over time, do not seem likely to have the same exponential improvement that we have seen in computing hardware. Again, though, we know that a machine with human dexterity is possible (because hands exist) so it seems inevitable that machines will eventually surpass us in these abilities as well.
There are, however, certain jobs that some people may be willing to continue to pay for a human to do, even if a robot can do it better in some sense. For example:
– producing handmade goods
– creating artwork whose value depends on whether it is an original
– some kinds of food preparation
– performance arts (acting, dancing, stand-up comedy, etc…)
– domestic service (personal servants like butlers, gardeners, etc)
– sports
– therapy
– hairdressing and the like
– massage
– certain aspects of medical care (a sense that someone cares)
– certain aspects of teaching (motivation, mentorship)
– certain aspects of war (decisions about when to use violent force)
– clergy work
– mortuary services
– some kinds of sales
– politics
For these kinds of tasks, having it done by a human is part of what is valued by some customers. If most other jobs can be automated, more jobs that fall in this category would be expected to be created, as a larger pool of workers is available to do them.
As more jobs are automated, what economic effects should we expect to see?
Around 1800, economist Jean-Baptiste Say argued that workers displaced by new technology would find work elsewhere once the market had had time to adjust. By the mid-1800s, a theory was in place that explored the economic effects of automation. In Das Kapital Marx would later dub it “the theory of compensation.” This includes additional employment in the capital goods sector, decrease in prices, new investments, and new products (the effect on wages is complicated). In general this is still the prevailing opinion of economists.
According to this theory, when workers are fired because their jobs are automated, this frees up capital which the owner will then use to hire other workers to do other jobs. Because of this, the number of workers hired doesn’t decrease because of automation. (Marx disagreed with this, saying that part of the capital would now be tied up in the machines). The theory also discussed other effects. Automation reduces the prices of goods, making them more affordable. It also reduces the prices of components, making new products viable. The companies making these goods make more profits, allowing them to expand and hire more workers.
Because of these effects, as long as the market has time to adjust, we shouldn’t expect to see increasing levels of unemployment up to the point where robots have taken all the jobs. Instead, new jobs for workers should be created until the entire employment pool is being utilized. This process can be expected to continue up until the point that all jobs can be done more cheaply by machine. As long as there exist skills humans can do more cheaply than machines, the number of jobs using those skills should increase until they absorb the entire available human labor pool.
Susskind 2018 concludes that in the future, automation will put downward pressure on wages, while increasing the amount earned by capital owners. We may already be seeing this effect in the United States: although GDP per capita and net productivity have increased consistently since the Great Depression, median wages have stagnated since 1975. This would be consistent with automation sending increases in productivity to the owners of capital rather than workers.
As more jobs are automated, the mean standard of living will improve, as the amount of value produced per-capita goes higher and higher. Even without raising tax rates or rates of giving to charity, the overall amount received will increase as more is produced at lower cost. Whether this leads to more people living on the dole or not is more a matter for political argument than for technological extrapolation.
Another effect might be shorter working hours– as more jobs become automated, the same number of people could be employed, but at fewer hours per week or more days of leave per year. Given the option, though, the preference of most workers at the moment is to work full-time (and for many workers, overtime) trading leisure time for additional income. For this to change would require both regulatory changes (part-time workers have different rules about benefits, for example) and cultural changes. It is not absurd, though: in Germany, for example, the average adult only works 1400 hours a year (26 hours a week) compared with 1900 hours a year (35 hours per week) in the U.S.
One might expect that as machines become more capable, more and more people will find themselves below the waterline, unable to find any job that AI can’t do better– perhaps those with the lowest IQ first, or something along those lines. To date, though, the capabilities of AI have not developed this way. Grandmaster chess and rapid calculation can’t be done by those with low IQ, but are simple for modern machines, and the inverse is also true– even very young children and those with a low IQ can perform recognition tasks in varying conditions that defeat even the best computer vision programs, for example. On the other hand, the number of routine jobs in the U.S. is an ever declining fraction of all jobs. If some constant fraction of people can only perform routine jobs, eventually some of them will be unable to find any job they can do, if current trends continue.
Suppose we reach the point where robots can do literally any job a human can do. What will happen to the economy?
A robot will likely never be cheap compared to other manufactured goods. Although future process innovations (such as advanced 3D printers or nanotech assemblers) may reduce the cost of building robots, they will also reduce the cost of manufacturing everything else, and robots must have large numbers of moving parts. This could mean that the number of robots will be limited, and this limited supply will drive up wages in jobs that the robots could otherwise do, if there were enough of them or they were cheap enough to produce.
Robots with human levels of ability, however, would be able to self-repair, extract natural resources, manufacture parts and create more robots without any human intervention. They would also be able to invent new ways to make money and employ other machines to achieve goals.
For any job, the machines in this scenario could do it better. People will still likely strive to purchase and direct factories, resource extraction, and robots for all purposes. Those doing this would still have a job of deciding how to direct the robots, acting as business owners. (Although one imagines running an AI as a manager to handle this kind of work as well.) There will also be people with earned or inherited wealth who don’t work, and welfare recipients who don’t work, but to what extent the economy will redistribute the wealth generated by this vastly expanded economy is a political question.
If we ever reach this point though, it is hard to make serious predictions because we don’t know what such machines would be like, and whether we would be able to maintain control of our economy and civilization at all (this breakdown of all models is the reason von Neumann called such an eventuality a singularity). If it becomes possible to create machines with human level intelligence and skills it will be possible for a little more money to create superhuman intelligence and skills, which will necessarily be hard for us mere humans to predict or control.
Conclusion
Even if nearly all currently existing jobs will eventually be automated, as we progress toward that point new jobs will continue to be created for humans, preventing the kind of mass unemployment or low wages that might be expected, as long as the market has time to adjust (which isn’t necessarily the case– if the pace of automating jobs were to speed up enough, we could still see a crisis.) However, once machines surpass human capabilities for a low enough price in all jobs, the entire economy will change and something else will take its place. What that is we can’t really say– our economic models break down, and the future becomes even more difficult to predict. Beyond this point, we don’t even know to what extent humans are still guiding the course of civilization, let alone how employment will work.
The economic gains that come from all this automation will flow primarily to those who own the machines. As they invest more, create new products, spend more, pay more taxes, and give more to charity, the general civilization will benefit, though some people will doubtless be unable to adapt and find new work and be worse off. How we choose to provide for those that can’t find work is something each democracy will need to continue to decide.
Further Reading
I found this article on Wired had some good points. One of them was that ‘job churn’ is at historic lows. That’s the rate of creation and destruction of jobs. You would expect that to go up if the rate of job automation was increasing.
The paper Automation and New Tasks: How Technology Displaces and Reinstates Labor provides a reasonable framework for estimating current and future effects of automation on labor. They conclude “if the origin of productivity growth in the future continues to be automation, the relative standing of labor, together with the task content of production, will decline. The creation of new tasks and other technologies raising the labor intensity of production and the labor share are vital for continued wage growth commensurate with productivity growth.”
If you are interested in specific predictions about dates for developments in AI, ML and robotics, MIT roboticist Rodney Brooks has a blogpost.
AI researcher Stuart Russell’s new book Human Compatible provides a nice introduction to current thinking about the future development of AI. It also contains some interesting ideas about creating artificial intelligence with open objective functions, so that the AI wants to please people but isn’t sure how best to do so.
So… will automation/AI, or will it not, lead to a general, sustained economic crisis within our lifetimes or for the foreseeable future? I read through this whole write-up, and I read your conclusion three times, and I still don’t know what either or both of you now believe is the answer to the yes-or-no question stated at the top…
This sounds like it’s leaning towards “probably not”.
Our conclusion was probably not, in the short term. Once AGI is imminent, however, there will be some kind of general crisis, I think.
Wouldn’t it be more accurate to describe it as a general “phase transition” to something entirely unpredictable? Indeed my assumption is that what determines whether it is a crisis or a cornucopia will most likely be determined by HOW society (or various societies) react to the changes and opportunities.
Also, “some people will doubtless be unable to adapt and find new work and be worse off.” May of course be true. But it is also possible for technology to lead to dramatic improvements in human adaptability. For example, assistive AI. Better education AI. And most importantly improved genetic engineering of babies.
I didn’t mean to imply that it would be negative. Perhaps crisis is the wrong word. I just meant that there would be a huge disruption of the status quo.
I’m pretty content with these things not ending with any firm conclusion. I could read an assumed conclusion into which layer of rebuttal/counter-rebuttal they always seem to stop at, but maybe that just indicates the person who spent more time writing.
I’m one of the authors. I say the answer is yes, but narrowly: artificial general intelligence is in the foreseeable future (whether it is in our lifetimes no one can say). When we hit that point, there will be a general economic crisis. So that makes the answer yes.
I went into this thinking that unemployment would gradually increase to 100% as we approached that point. I was convinced by our conversation that the curve will be more sigmoidal in shape, with the usual ups and downs overlayed.
Did you and your coauthor give any consideration to whether comparative advantage will have any effect, even in a world of strong AI/AGI?
Elaborating on the question: http://www.jonboguth.com/our-glorious-future-more-robots-more-human-flourishing/
Can you explain your reasoning? I don’t get it from the essay.
E.g. Amazon is nowhere near AGI, yet it did cause many brick and mortar shops lay off their workers and shut down. Did all these workers find employment at Amazon fulfillment centers, or at Walmart or some equivalent employer? Did they #LearnToCode and get fancy jobs in tech? The Theory of Compensation assumes that all businesses are labor-limited and workers are perfectly interchangeable commodities, which seems to be empirically false.
Unemployment is at a record low in the US, so it seems reasonable to assume that most of those people were able to find other work somewhere.
But this is confounded by Trump’s anti-immigration and protectionist policies.
Seems unlikely. I don’t think either of those things has been shown to increase employment, and foreign-born as a % of the population is at a >100 year high
So, if it’s really important how fast we expect automation to change the job market, and how fast we expect labor supply to be able to respond, what sort of sources might tell us about these two timescales?
Presumably there’s economics work on the second part of this question – maybe even experimental economics work. A quick google search turns up this paper – it might be if interest, and maybe it references other papers in a similar vein.
The first part of the question is a bit more case-by-case, but it might be worth considering some scenarios. The most extreme near-term shock that we might consider is the automation of the trucking (and general vehicle-driving) industry. Suppose that someone invents a product that can do the job of 80% of truckers – how should we expect the cost of this product to change with time (e.g. compare to the price of other automation solutions after initial development), and how should we expect adoption to progress, given some assumptions about cost (e.g. compare to adoption of other production technologies)?
Since my expertise is more in AI than economics, that’s what I focused on. I would like to read the essay you are picturing, too. 🙂
I think automation is already a big problem, will be a larger problem going forward, and I didn’t feel like any of the reasons why are being represented in this AC. Consequently, I don’t really consider this as evidence in either direction. To be more specific, the things I was hoping to see here were
– looking at labor force participation rather than unemployment. How did it change so far, how did it change in the past, what can we expect for the future?
– trying to quantify the economic instability caused by previous instances of automation, probably the industrial revolution is most relevant here, trying to quantify how large the current wave of automation is compared to that, and then trying to estimate how bad things will be this time
– trying to quantify what happens to most people who are pushed out of work
– evaluating how plausible the claim is that automation was the primary cause of Donald Trump’s election and trying to quantify the results of that
None of this is necessary, if the AC instead came up with even more meaningful ways of approaching the question, but I really don’t think it did. Essentially, the section “As more jobs are automated, what economic effects should we expect to see?” gives one standard argument and then leaves it there.
I grew up and live in the suburbs of Washington DC. My maternal Grandfather did too. He worked in the federal government mostly in the 60s and the 70s. He has talked about how many unskilled jobs there were available in the Federal Government (by far the primary industry of this place) as couriers, copy clerks, transcriptionists and the like. These jobs basically don’t exist any more.
This is, in a large part, due to word processing, e-mail, and other similar computer advancements. This has undoubtedly increased the productivity of the federal workforce and freed up tax money to be
wastedspent elsewhere instead ofwastingspending it on the machinery of the bureaucracy. I’m not saying this is a bad thing but it has had amazingly profound effects on the local culture and economy.The unemployment patterns in DC are interesting (this pattern in Federal employment is found mentioned in the second paragraph of the last section labeled “Moving Forward”). Anecdotally, government officials, politicians, and lobbying firms are hiring a lot of people who live elsewhere to move to DC to work (it helps that if someone wants to work with anything attached to the Federal Government of national politics that one would want to come to DC) rather then train people who already live here. Undoubtedly technology has also impacted this trend whether you think this trend is a good thing or a bad thing.
This seams like a case where automation has cost people jobs in a way which has left generational effects. Perhaps the market hasn’t stabilized yet (to markets ever stabilize?) but the existence of generational effects renders the issue of automation of jobs a pertinent one. Certaintly one can have the position that nothing special should be done and people should be left to adapt or not but one can also have the position that something should be actively done to mitigate the ill effects of automation and/or help people to adjust.
Depending on how one defines a “general sustained crisis,” one could argue that automation already has contributed to one for some of the neighborhoods in Washington DC. Certainty it has been culture changing and this is shouldn’t be looked at in isolation. If DC had a competent educational system, perhaps there wouldn’t be such a crisis, for example. Never the less this is evidence that the issue is not about the future but rather about the present.
A very interesting discussion, and a critical one. And one that challengers right-wingers like me, who worry about the social aspects of mass permanent unemployment, even if we could and did give everyone an automatic “social wage”. Would our unemployed-but-affluent citizens be using their leisure to study metaphysics, or methamphetamines?
One thought: progress in electronics and software is paralleled by progress in genetic engineering. Perhaps in two or three generations, the children growing up with very advanced AI, will have been genetically engineered to have very advanced Natural Intelligence (plus all the other biologically-influenced socially-desirable traits).
Then it’s goodbye to the Kingdom of Necessity and off to explore the Kingdom of Freedom.
How many retired people use methamphetamine? How many idle rich people? It strikes me it’s mainly the “employed (or should be)” class that uses it.
Even with advanced 3D printers making every actual product, you’ll need lots of robots to mine the raw materials and work in factories, processing the feed material to ready it for the printers.
You might, but those would likely be robots in space, mining asteroids. Which isn’t a job we can send people to do period, making the impact on jobs moot.
I’d like to see the evidence for the preference of most workers to work at least 40 hours a week.
It seems to me that, if I simplify it, we have two kinds of workers. First, those who need to work 40 hours a week to feed their family. Okay, these seem to have genuine preference for not seeing their kids starve.
But for the other group, those who could live on half of their actual income, it seems to me that we have a cultural norm of pretending that you enjoy your job so much that you would do it even for free; and that you are asking for a salary merely to cover your expenses and because it is the usual thing to do, not because you want to be compensated for the painful loss of your free time which is the only time when most things personally valuable to you can happen. (All that unprofessional stuff like being with your family, with your friends who are not your colleagues, having a hobby that is unrelated to your daily job, et cetera.) Because we all find our greatest pleasure at making our bosses even richer, and spending a large fraction of our lives at meetings. We all love “challenges”, especially the jobs that suck, artificial deadlines that lead to stress and overtime work, following arbitrary rules, playing the zero-sum game of having our productivity constantly compared with our colleagues, etc. At least we all pretend to — because if you don’t, the job will instead go to the guy sitting next to you who is better at saying all the right words at the interview.
The problem with asking for a part-time job is that it ruins this pretense. It makes it hard to deny that there is something in your life that you’d prefer to do, rather than spend most of your daylight time (including commute and lunch, pretty much all of it, especially in winter) sitting in the office. You better come with a really good excuse why you’d prefer to work full time but regrettably you are not able to: for example, if you are a woman and have small kids. But if you don’t have an excuse ready, for example because you are a guy, or because you are childless, then asking for a part-time job simply signals you are not that much in love with your job. Why would anyone hire someone with that kind of attitude? (Unless they are desperate, of course, but then you probably can’t expect high salary.)
Speaking for myself, it’s not that I prefer to work 40 hours a day, but I simply can’t find a job where I would work 80% of time for 80% of my usual salary, or 50% of time for 50% of my usual salary. And I tried for years. (The closest deal I was offered was to work 80% of time for 50% of my usual salary, and even then the potential employer felt they were doing me a great favor.) So when I work full-time, it’s not because I believe that 40 hours a week is the optimal amount of work, but simply to avoid the “weirdness penalty”, which is very steep. In a country where working e.g. 30 hours a week (for 3/4 of the 40-hour salary) would be normal, i.e. no one would stare incredulously at you during the interview, I would take that option in a heartbeat.
I wonder how many people out there are like me. Having the “revealed preference” of working 40 hours a week, while bitching about the “weirdness penalty” associated with our true preferences.
(If you are the type of person who believes that 40 hours is optimal for them, imagine living in a country where 50 hours is the norm, and anyone who asks for 40 is considered a weirdo, and most employers prefer not to hire weirdoes. So you work 50 hours a week, and you pretend to like it, because the alternative is having no job, or finding someone who pays you peanuts for 40 hours because they feel super charitable for giving a chance to a weirdo. Then you would read the articles about how people prefer to work 50 hours, and realize that the authors also consider you as another piece of evidence for this “preference”.)
I didn’t mean to talk about what people really desire in their hearts. I just meant in the economic sense of revealed preferences: the majority of hourly workers, when given the choice to work more hours for more pay, choose to do so. What their motivation is I don’t know.
Probably because hourly workers are generally paid at a low enough rate that they essentially have to work 40 hours or more just to afford basic life expenses.
In countries like Germany and France, where working 40 hours a week isn’t the norm (partially due to culture, and partially due to regulation) , it doesn’t seem like there is a big push from employees to work up to that amount.
I just meant in the economic sense of revealed preferences: the majority of hourly workers, when given the choice to work more hours for more pay, choose to do so.
Whenever I see definitions like this, I am honestly curious as to whether the economists’ theoretical explanation for slaves continuing to work under the control of their owners was that it was a ‘revealed preference’ (I mean, they could always have decided they preferred to be murdered by the overseer for refusing to work!)
Economic theories that admit they don’t take into account coercion or necessity are great theories, but we really are in Carroll’s Wonderland when it’s “workers prefer to work X hours but we don’t know why workers prefer this, we have no idea why workers work, it is just something they do for unknown and unknowable reasons”.
Some people may love their jobs so much they genuinely would do the work for free (if they had other sources of income to prevent them starving to death); most people do more work because it gets them more money. Why would people want more money? Well, if economists genuinely pretend they can’t come up with a few reasons why, then I don’t see what we can do about that.
Of course it is: it is a revealed preference for being a slave over being murdered. Naturally, it doesn’t show that they have a preference for being a slave over living as a non-slave.
I am not sure where you got the idea that economists “can’t come up with a few reasons why [people want money]”. Obviously there are many such reasons, but I don’t see how it invalidates the concept of revealed preferences.
The question was “will people work fewer hours in the future?” To answer that question, I looked into whether people who are given the choice work fewer hours now. What I found was that hourly workers argue over who gets to work available overtime hours because they need the money more than the time. Until that situation changes, I expect people won’t choose to earn less money to have more time off. I’m not making a value judgement, just a prediction.
Viliam,
You wrote “playing the zero-sum game of having our productivity constantly compared with our colleagues”. Seems to me that this is pretty much the antithesis of a zero sum game. It is the poster child for positive sum games. Workers are competing constructively to increase productivity, with productivity gains going to consumers* (and workers are also consumers, so they benefit indirectly from the general dynamic).
* The productivity gains do not go to employers long term due to a similar constructive competition which drives down profits to the risk adjusted rate of capital. However the employers are consumers as well, so they gain too.
So, the software industry?
Count me as one.
First, I believe that first group (the ones working 40+ hours a week to feed their families) is a distinct majority. Second, I believe you’re largely ignoring the portion of the second group who plans to be in the first one, which is probably a majority of that group. The desire to work 40+ hours isn’t so much about the present, but the desired future: you want to have the excess to be able to support a family when that time comes. Yes, there is some percentage of people who will never be in that group, but they’re a minority and workplaces (and society in general) has a cost for that kind of oddity which doesn’t conform.
In essence, the top-down approach in Europe is likely the only way to actually reduce work expectations because of the coordination problems inherent in getting large numbers of people to defect for (for example) 80% of the time for 80% of the salary.
Really? To FEED their families?
You might have more success negotiating for more vacation time rather than an explicitly shorter workweek. Say, if you could manage to get 6 weeks vacation instead of 2 weeks and draw 90% of your regular compensation. That’s probably a difficult arrangement too, but another tack to pursue at least.
Also, keep in mind that your employer has a lot of fixed costs in hiring you (e.g. health insurance, other benefits), so paying you 75% of normal compensation might equate to, say, 60% of your normal salary (and 100% of your non-salary costs).
Well said. I agree with pretty much all of this. I would love to work 20 hours a week, and I could comfortably survive and save on it if I got anything like the hourly rate I could expect on a full-time job with my skills. But the jobs just don’t exist. If I want to work 20 hours a week, I’m doing something (a) less interesting and (b) much lower pay. It seems to me that one of the great challenges of the 21st century is to get us all working less hard and having a bit of leisure time for once.
And I’m in Europe. I understand it’s much worse in the US.
“The economic gains that come from all this automation will flow primarily to those who own the machines.”
I am not following the economic logic of this. As I understand it, those who will own the machines will engage in economic competition with others owning similar machines. Thus the risk adjusted return on the investment will be attracted the going rate. IOW, variable profits will be similar to current profits over the longer term.
Thus the benefits will flow primarily to consumers, as has always been the case. Certainly successful entrepreneurs will make big bucks, but there is no way in advance, of course, to know which will or won’t be successful.
Or am I missing something?
Your first paragraph doesn’t appear to contradict the quote. Non-robot investments may make similar returns to robot investments, but those returns will not be due to automation. Furthermore, people who expect more automation than the market does have an incentive to invest in robots, since the more automation there is, the better an investment robots will turn out to be relative to other investments.
Just to clarify, I agree that competing investments of all types will drive down profits toward the longer term risk adjusted cost of capital.
Producers won’t make higher or lower rates of return long term than they have tended to in the past. The benefits will go to consumers, as has always been the case in relatively free markets. We are all consumers, thus we will all benefit as consumers. The relative gains will be distributed with extremely large variance of course.
As I understand it, what usually drives down large profits is limits to the expandability of organizations: eventually it is impossible to continue making robots and sell them at the same price for the same profit. The stock market allows robot stocks to expand until this happens. So in a sense the market “drives down profits”, but in a more direct sense the profits are just being driven down by fundamental economic limits.
For the record, I don’t really know how well off various groups will end up in an automation scenario, and I am not really sure how to think about that question.
You may be confused, it sounds like you’re describing the profits of producing and selling robots, whereas swami (and the op) are discussing profit derived from replacing wage-labor with robots in existing production chains: I.E. the share of sale price that used to pay for worker wages over the cost of running robots to perform those tasks. Op says those profits accrue to business owners, swarmi contends they will (at least over the long run) be used to reduce the sale price to undercut competing firms.
Also, there’s no reason to imagine robots would be different than cars or smartphones in that market saturation isn’t the limit of the market. People who already own cars and smartphones are the primary market for those same products.
@Nicholas: I think you are referring to when I said “eventually it is impossible to continue making robots and sell them at the same price for the same profit”. This was meant as a simplifying example, but perhaps it caused more confusion that it was worth. For another industry that uses robots you could say “eventually it is impossible to continue selling robot-based goods and services at the same price for the same profit”. (Maybe I should also clarify that “eventually” means “if these industries become too large in comparison to the number of potential customers”, which might not ever happen but which will happen if there is too large long-term growth.)
I wasn’t claiming that the limit comes from saturation, but only that it exists, and that the industry will expand until it reaches the limit. In other words, if an industry is expanding faster than the market in an unlimited way, then eventually it will just take over the market and the market will expand at the new growth rate. I could be wrong, but if so I would want to know what else could be the mechanism for the market to “[drive] down profits” of an overprofitable industry.
Competition is what typically drives down the profits of an industry with excess profits (rent). Other investors enter the market to capture the higher returns, driving down the margins.
@Cliff: Ah right, I probably should have thought of that.
+1
I think you probably know more about this aspect of it than I (the co-author) do.
https://www.militantfuturist.com/will-future-technologies-end-capitalism-no/
I mentioned UBI in the introduction and agree that’s one way things could go, but you can imagine things working out in several ways that don’t include UBI. UBI does seem to me like an elegant solution to the situation someday, when much more wealth is being created and it would be less of a burden on those working to provide it with their taxes. I think my coauthor Erusian would disagree, though.
“Systems that have lifelong learning, that continue to grow with experience, are still very rare.”
To be fair, this is also true of most humans, though to a lesser degree and over longer timeframes.
I agree with the conclusion – there will probably eventually be a crisis, but we don’t know when it will happen, and how bad it will be is primarily a political question. That said, I wonder about two things:
1. Do you think people will continue to prefer humans to do certain jobs once there are automated systems that work “better”?
2. You mention that most benefits of automation accrue to the owners of automated systems. This is obviously true today, but as with your point about political questions – what are the odds you assign to people allowing this to remain true in the future? If *any* human or set of humans is in control at that point, I hope other humans will be able to prevent them from seizing complete control of the world a la Manna, but I’m not really sure of that.
My answer to 1 is yes, I think some people will prefer humans to do all the types of jobs I listed even when robots can do them better. For many of these things, what you are paying for is a kind of story that you tell yourself, and a relationship between yourself and the person doing the selling. Artwork originals are a good example. Original artwork sells for much more than reproductions, even when those reproductions are very high quality or even indistinguishable forgeries. Robot sports are currently less popular than human sports, but it’s not because robots are slower or less accurate than people. Some people like chatting with their hairdresser about their life, and knowing that a robot was only pretending to be interested wouldn’t feel the same. They might want a human to give a eulogy, even if it isn’t perfect, because they feel it is a spiritual thing and that robots don’t have souls. This is the kind of thing I’m thinking of. Other people may become convinced that robots are really people too, and not feel that way.
Every human hairdresser is also just pretending to be interested. Maybe they just are more effective at pretending.
My answer to 2 is I don’t know. I thought I could predict some things about politics but then Trump got elected and I’ve doubted my ability to predict anything political ever since.
This feels pretty central to the whole subject, and it’s not discussed very much. It seems plausible to me that we could have several decades during which robots can’t do every job, but there simply aren’t enough jobs for people who aren’t good at creative thinking, problem-solving, or whatever you want to call it.
A lot of AI tech is fairly generalizable. The same machine vision tech that enables self-driving vehicles, for instance, is going to make machines better at taking over for factory workers, warehouse workers, security guards, doormen, store clerks (the single most common job in the US by some counts), etc. So when you lose your job as a truck driver to a few dozen dollars worth of computing hardware in 2028, you probably won’t have much luck finding work doing any of those things either.
The industrial revolution created factory jobs even as the mechanization of agriculture eliminated farm jobs. Almost everyone capable of farm labor was also capable of factory labor. Will the AI revolution have this property? What’s the equivalent of factory work here? As the article notes, the number of ‘routine’ jobs is already declining.
We should be especially concerned about this, because having, say, 30% of the population be permanently unemployable, while the still-employable people and those who own the automated means of production see rapidly improving standards of living, sounds like a worst-case scenario for social unrest. Unlike in a scenario where everyone or almost everyone becomes unemployable, in this scenario the (still-employable) majority won’t be very motivated to vote for redistribution to the unemployable, leaving the unemployable with little recourse through democratic institutions.
Take a look at the lives of the citizens of the Persian Gulf Petrostates to see what a modern first world country would look like with no labor shortages. Everyone with kids has a nanny or three to take care of them. Everyone has their home cleaned by a service. People get a large fraction of their meals delivered, resulting in soaring obesity rates. And that’s just the jobs that directly interact with the lives of the average citizen – the same countries also employ enormous numbers of low-skilled men in construction and so on.
When all members of the Profesional Managerial Class have the same access to services as a Qatari citizen or an A-List Celebrity with an entourage, then maybe we can start worrying about technological employment for the lower classes. But it doesn’t look like we’re anywhere near there yet.
https://www.militantfuturist.com/the-future-is-already-here-its-just-not-very-evenly-distributed/
Am I the only one questioning whether a robot can provide “quality” childcare?
The Persian Gulf Petrostates are very different in culture and society from a first world country without labor shortages. Yes they have crazy high obesity rates. But in Western countries it’s specifically the lower classes who binge on delivered junk food and get obese. The upper classes, while they may have ordered burger and fries 50 years ago, now are more likely to go for sushi and salad, and to go to the gym and generally keep a healthy weight. Also the Persian Gulf climate makes it nearly impossible to walk or exercise outside, which is not generally the case in Western countries.
“A lot of AI tech is fairly generalizable.”
Except, in practice AI that is deployed to production is almost always part of a larger pipeline anyway, and you end up with the same debugging/maintenance/testing costs of any engineered system. Self driving cars are (from my understanding, and in most cases) are not trained end-to-end with generic techniques, they are highly engineered systems with perception, control, rules, and more. I am speaking as a PhD student working in AI here by the way — it seems to me many technical people who have not gotten their hands dirty with modern day way overrestimate how easy it is to apply to any given new problem.
Beyond the near term and the effect of automation pace, this discussion is just fantasizing. The point where this discussion makes no sense is when machines surpass human abilities. To try to think about that we need to understand the difference between life (living) and machine but we seem to be determined to not understand life and it’s purpose.
My notion is very likely completely uninteresting/unacceptable to the majority of readers here and I believe b/c there is no reasonable counter.
BTW my view has nothing to do with god or spirituality or intelligent design.
Do we truly think the human brain is an Exaflop? That’s pretty insane if true. This puts Moravec’s estimates at more than 6 orders of magnitude too low!
The estimates I found spanned many of orders of magnitude, and my estimate was somewhere in the middle. If we can simulate a working human brain by simulating the neurons at the resolution I suggested, then I’m sure with optimizations for the hardware it could be done with far less than an exaflop. However, some people think that a lot more processing goes on inside a neuron that is important. If you’ve ever seen an amoeba crawl around capturing food, avoiding hazards and so forth, a single neuron is probably as clever as that. Each synapse, which I have being represented via ten thousand flops per second, has a lot of complexity in its own right (as you can see in microscopic images of them) and the dendrites themselves have been shown to have some computing capacity. So, it could be more and it could be less.
Here is a good summary of the various estimates that are out there.
Whenever I see a chart with such a clear point-in-time separation like “disconnect between productivity and a typical worker’s compensation”, I get a bit suspicious.
This chart suggests that some change occurred in 1975, and after that everything proceeded in the new paradigm without further change (i.e. in this case, a new rate-of-change for worker compensation was established). But that doesn’t really fit the narrative of “gradually increasing automation”, which I’d expect to apply a deceleration rather than a one-off change in velocity.
Another possible explanation is that this is a statistical artifact due to a change in how the datasets are being reported.
I’m curious if anyone is familiar enough with this data to shed more light on this part of the argument.
Scott did a whole post on this. It definitely should at least include a caveat that the chart is hotly contested. Most people find that when using the appropriate inflation measures, etc., compensation has increased a lot more than 9% in real terms since 1975
Do we know how much of the lower hours worked per year in Germany is due to people actually being more likely to work <40 hour weeks when on the job, vs stay employed less of the year and/or take more vacation, longer parental leave etc? Either would decrease hours worked per year, but the burstiness, types of coordination challenge created, and general feeling of "do I usually work full time" are quite different in the two scenarios.
In Germany, employees are entitled to a minimum of 30 days of paid leave a year (10 paid holidays, 20 days of vacation leave). In the U.S. that minimum number is 0 (the average is 10). That’s where most of the discrepancy comes from, I think.
You mean 10 days of vacation leave? Most if not all get federal holidays as well, which must be at least 10 days/yr.
I don’t want to take away from the effort invested in this collaboration (the authors are to be commended), but I do have three issues with the analysis:
First, figure A is graphing apples and oranges. It compares average hourly compensation of production/non-supervisory workers to the productivity of the entire economy.
Let say for some reason, say technological advancement, after 1975, the effect of productivity increases changed the composition of the workforce. So less hourly factory workers than before over time and more supervisory/salaried engineers/service workers. Why would we expect productivity gains to accrue solely to the remaining hourly factory workers rather than the newly created job-types which are key to the actual productivity increases?
To do a real comparison, you’d either need a number for productivity gains in the sub-sector of jobs you’re reporting wages of, or else compare overall productivity gains to overall compensation, not just a subset. Sure, the two numbers tracked better when the workforce composition meant that factories employing mostly hourly workers was where most of the jobs were, but that changed with the computer and information revolutions.
Second, (and this is a nitpick) this final line makes a lot of assumptions:
There was no demonstration that those who can’t find work will increase (the empirical track record says otherwise) and even if it did, it doesn’t follow that “each democracy” has to decide that. Individuals or voluntary groups can decide what they want to do. Everything doesn’t automatically need to be decided politically.
Third, the rest of the final paragraph contains an internal contradiction:
The economic gains flow primarily to those who own the machines…. and then to general civilization. Well, which is it in their opinion? I think they’re missing the time factor in their description, i.e. initially a disproportionate amount of the gains (I’m not even sure you can say the majority) will flow to the first movers who own the first sets of machines, but with competition they will quickly only receive around the same market-wide returns to capital, because if they are getting much more than that, it will induce more capital to enter the same markets until there is no longer a big incentive to do so. This is a basic economic reality of capital markets.
The new equilibrium of capital returns may end up a little higher, but it will also be much broader than they seem to assume.
Also, a lot of capital in the U.S. is provided by a wide swath of individuals. Most middle-class and above individuals have investments via 401ks, etc… All those people will be among the group to benefit if the returns on capital increase for everyone because of automation.
In addition, there is a very short time frame between wealth-increasing innovations (of which automation is a big one) and when that wealth spreads out into society. Hence the infamous cost disease, the poor in one country being richer than the rich either in the past or in another poorer country, and similar effects. Sure, some people individually benefit more because of their more direct-relationship to the actual wealth-building, but it doesn’t take long for others to benefit as well, because (and this is the biggest point not considered in that final paragraph, so I’m going to emphasize it):
Users of automation only benefit from it in direct proportion to how well they can use that automation to benefit the consumers of the products of that automation.
Therefore, the consumers also benefit from increased wealth as a result.
Those interested in this may also dig a thing I helped edit and write, Job loss due to AI — How bad is it going to be?. It covers some of the same points but is a bit more in-depth (many more sources). The TLDR is similar:
“In this article, we summarize the findings of recent studies and argue that the impact of AI on jobs in the near future will (most likely) not be significantly more disruptive than the impact of automation has been in the past. The number of jobs created by AI and automation in general will likely outnumber the jobs displaced. However, because workers of different countries, sectors, and income-levels will be affected differently, we still need governments to proactively enact prescriptive labor policies to facilitate smooth workforce transitions.”
That’s not the constraint.
The constraint is *how fast you can retrain displaced workers*. As we continue to learn how to automate things we decrease the time between the creation of a new job/skill and the time when is automated away. Thus we will need to retrain workers at faster rates in the future.
We will have to teach people to learn faster, and teach teachers how to teach faster, but you’ll run up against the limits of humanity.
This doesn’t really address the relevant questions, which in my mind are:
Will we actually see certain jobs automated away?
If so, how fast will this happen/which jobs will be affected?
Assuming some people are automated out of their jobs, will they in general
1. be retrained?
2. be un/severely under-employed, but that will be OK because goods will be cheaper?
3. be un/underemployed, but that will be OK because UBI or whatever?
4. be horribly poor and miserable?
I don’t think extrapolating from option 1 being common historically is in general reasonable, unless you have theoretical reasons to support that which don’t seem to be given here.
This write-up’s points seem cogent and well-supported and I was persuaded that automation is unlikely to cause economic crisis. But I’m probably an easy audience for this because I believe the pace of automation will be slower than expected. (Main reason: Designing & building robots is still very expensive relative to human labor, we shouldn’t expect *that* to get much cheaper because it’s skill-intensive and capital-intensive, and if it gets significantly automated then there are good reasons to believe we’re close to the Singularity anyway.) My main reservations are:
– I can’t tell how energetically the other collaborator (Doug Summers-Stay) was opposing this conclusion
– Not much attempt at rigorous or numeric analysis, even to the point of estimating probabilities or laying out scenarios
– Narrow definition of “economic crisis” that boils down to framing it almost entirely in terms of GDP. Structural unemployment due to rapid automation and out-of-control disparity of incomes can’t be dismissed as forms of economic crisis.
This was a shorter entry that reached clear, solid conclusions with well-written arguments and lots of pretty graphs. Maybe I’m looking a gift horse in the mouth here, since I dinged other entries for being wishy-washy, but unless it’s clear what arguments and counter-arguments were considered I have trouble distinguishing this from a FAQ-type document that takes a specific point of view. The reliance on analogies and thought experiments exacerbated this problem. So I didn’t get the kind of content from this write-up that would make the adversarial-collaboration aspect worthwhile.
I rate this collaboration as a 7/10 (no particular scale or judgment implied; this is just for my own reference). As always, many thanks to the authors for putting in the work to create this.
Minor pedantic point: “So we know from experience that employment is able to adapt to extensive automation. However, these changes took place over millennia, centuries and decades respectively.”
This is an incorrect use of “respectively”. It should be used when there is a list of things that each corresponds to a particular one in another list of things, not just plonked on the end of a list of things. It tripped me up when reading it, because I assumed I’d misread something, and that you were saying one type of change took place over millennia, another took place over centuries, and a third took place over decades. So I went back and read the sentence twice more to clarify. The sentence would make perfect sense if you just removed the word “respectively”.
I think that nobody is talking about a big issue – that the growth of the ratio [total production]/[total employment] is at historically LOWER levels, exactly the opposite of what should be expected if there is a big automation going on.
My guess is that most of the automation are occurring not at production level, but at consumer level – not machines used in factories or even offices, but more gadgets that people use at home (like smartphones).
The supposed “stagnation” of income since 1975 is one of those Big Lies.
Income has approximately doubled in real terms since 1975.
This is obvious if you look at literally anything. Houses have gotten almost 1000 square feet bigger, and have vastly more and better creature comforts – AC, central heating, no lead paint, better insulation, wired for internet access, ect. And at the same time, cars have gotten vastly better, people have twice as many TVs per household, and we have vast numbers of things that people lacked in 1975 – better health care, computers, cell phones, smart phones, internet access, VHS, DVDs, video on demand, satellite TV, streaming, ect.
Once you realize that is obviously all impossible if people haven’t gotten any richer, you realize that the actual answer is that the claims are just the result of pure manipulation by monstrously evil people for political gain.
CPI (the inflationary statistic they’re using in the graph) itself is pretty useless; it overestimates inflation by more than 1% per year, which is a cumulative effect which obviously is going to make the number wrong. The opposition to moving to a more accurate inflationary statistic is political; beyond the fact that it would undermine all the claims of the far left, it would also mean that various agencies and other programs have had their funding grow much faster than inflation, as the “automatic” inflationary adjustments have actually given people more money than they should have.
But it also ignores the fact that non-wage compensation (which is also paid for by employers) has grown vastly since 1975 – in fact, it has grown much faster than wages have.
Once you realize that the “disconnect” graph is just a flat-out lie, you stop having to make excuses for why people have so much more stuff and such a higher standard of living despite their income supposedly not having grown.