Take Both Econ, Techies Seriously

Martin Ford, software CEO and author of a new (bad) book on how automation may destroy our economy, is right about one thing:

Among people who work in the field of computer technology, it is fairly routine to speculate about the likelihood that computers will someday approach, or possibly even exceed, human beings in general capability and intelligence. … While technologists are actively thinking about, and writing books about, intelligent machines, the idea that technology will ever truly replace a large fraction of the human workforce and lead to permanent, structural unemployment is, for the majority of economists, almost unthinkable. For mainstream economists, at least in the long run, technological advancement always leads to more prosperity and more jobs.

Yes, techies agree on the long term plausibility of machines doing almost all jobs at a cost below human subsistence wages, thereby gaining almost all income, while economists ignore this scenario.  E.g., Tyler Cowen:

In the longer run … computers will free up lots of human labor — but in the meantime it will have drastic implications for income redistribution, across both individuals and across economic sectors. … Robin Hanson believes we are headed back toward a Malthusian equilibrium; in contrast I believe that machines will never outcompete humans across the board.

Arnold Kling agreed:

I agree that Singularians are far too optimistic about artificial intelligence. It is a variation of the “fatal conceit” problem. Most of human intelligence is tacit knowledge, consisting of elaborate metaphors that are originally acquired from sensory experience. Artificial intelligence is an attempt to arrive at the same point through top-down design. … Computers and robots will be economically significant but not paradigm-shifting.

Economists should listen more to techies on what techs will be feasible at what costs, but techies should also listen more to economists on the social implications of tech costs.  Alas, just as economists prefer to rely on their intuitive folk tech forecasts, techies prefer to rely instead on their intuitive folk economics.  E.g., Martin Ford’s misguided intuitions:

At some point … machines will be able to do the jobs of a large percentage of the “average” people in our population, and these people will not be able to find new jobs. … This tremendous selection of products, and also services, which we now take for granted, is unprecedented in human history. … All these products owe their existence to the mass market. …When a business creates products or services at high volume, it realizes economies of scale, and that, of course, results in lower prices. In addition, however, high volume production also makes it possible for the business to adopt statistical quality control techniques and to improve overall consistency and precision in the production process. …

As jobs are eliminated, … some of this wealth is then transferred to the owners. … The distribution of income [becomes] more concentrated. … The demand for the mass market products that drive our economy depend much more on the number of potential customers than on the wealth of any particular customer. You are not going to be able to sell 40 cell phones to one person, no matter how wealthy they are. … Continuing to save money as automation slowly eliminates some of their remaining workers, this is not enough to make up for the reduction in sales. …

If … a slave (or automation-based) economy is destined to undergo continuing decline, how is it that the slave states were able to maintain stability for so long?  The answer lies in the fact that the South was primarily an export economy. …

Once full automation penetrates the job market to a substantial degree, an economy driven by mass-market production must ultimately go into decline. The reason for this is simply that, when we consider the market as a whole, the people who rely on jobs for their income are the same individuals who buy the products produced. …  Making a few people richer will not make up for losing a large number of potential customers.

Ford’s mass-market theory of production is nothing like standard economic theory.  Sure high income inequality might be ethically bad, and threaten political instability, but it does not at all threaten economic collapse – producers can focus on giving the rich what they want, and innovation and growth is just as feasible for elite products as for mass products.

When I informed Ford than an economist (me) had taken his tech assumption seriously, he was not moved:

My immediate reaction to this is that economic growth at any level—let alone of an order of magnitude beyond what we are accustomed to—is fundamentally incompatible with wages that are falling dramatically for the vast majority of workers. … If wages fall dramatically, then consumption must likewise fall because the majority of personal consumption is supported by wage income. … Hanson … seems to be saying that if consumers have an ownership interest in the economy of the future, then the resulting investment income will be sufficient to make up for the precipitous decline in wages. …

Dr. Hanson seems to be assuming that the stock market (and other productive assets) would increase dramatically in value … The problem I see with this is that, according to modern financial theory, … asset values are defined by investors’ expectations for the future cash flows. … It seems clear to me that, in the midst of across the board job automation and plunging consumer demand, those future cash flows would be looking pretty minimal.

Ford seems to be confusing capital as stock vs. flow.  The fraction of production that is given to capital vs. labor depends on the marginal productivity of capital, times the quantity of capital, vs. the marginal productivity of labor, times the quantity of labor.  If capital and labor are the only owned factors of production, then if the fraction of income going to labor falls, the fraction of income going to capital must rise.  That income flow goes to capital regardless of what assets are used to represent the stock of capital.  Instead of trying to reason about what sets stock prices, it is simpler to imagine firms renting capital from capital owners, paying them according to capital’s marginal productivity.

The standard views of techies about what techs will be feasible might be wrong, and the standard views of economists of how to forecast tech consequences might be wrong.  And it is fine for contrarians to try to persuade specialists they are in error, though contrarians would be wise to at least understand the standard view before trying to overturn it.  But surely what the world needs first and foremost is to see and take seriously the simple combination of the standard views on such important topics.

Added 8a 29Aug: Ford responds by just repeating his claim:

Machines could go from performing 25% of jobs to 75% within four years. … Let’s try to imagine an actual scenario: … Businesses rapidly deploy the technology and the unemployment rate rises. … As consumer spending falls, businesses either adapt by laying off workers and automating still more jobs, or they fail entirely. …. Mortgage defaults are now at an unprecedented level. … How does the financial system and the overall economy survive that in the short run? I may be an economic rube, but I don’t get it. … Production is equal to consumption plus investment, and those are both going to be in free fall, given a scenario like the one above.

Yes, a sudden unanticipated change would be disruptive, but no disruption does not imply falling production.  Ford needs to learn some economics, or listen to some economists.

GD Star Rating
Tagged as: , ,
Trackback URL:
  • “If … a slave (or automation-based) economy is destined to undergo continuing decline, how is it that the slave states were able to maintain stability for so long? The answer lies in the fact that the South was primarily an export economy. …”

    BINGO!! All growth occurs from the outside in.

  • Peter

    If machines start replacing humans, doesn’t that mean that the humans can just sit back and enjoy the fruits of the robot’s labors? Won’t we all enjoy the benefits of being slave owners, whether or not we actually work for income? What I mean is, shouldn’t our absolute standard of living still increase even if we don’t necessarily get paid to work? Isn’t that the dream future: robots do all the work while we sit back and drink margaritas?

    • anon

      I think this is broadly correct. Of course, this requires people to earn some rents, but redistributing rent is much easier than redistributing wage/capital income.

    • Dan

      Hmm, yes but how do you get the ownership stake in the first place… the masses have almost automatic purchasing power because they can find almost immediate employment. How are the masses going to get capital to buy the stakes in the first place if they are “not needed” at all. Most people cannot or don’t want to own significant productive capital and to get a return on productive capital you need the masses, to consume consume consume…

      • anon

        See above; the masses can earn rent from land and natural resources, which is comparatively easy to redistribute. (For instance, the state of Alaska distributes natural resource rent to its citizens, and some local governments e.g. Hong Kong and Singapore are largely funded by urban land rents.)

      • “How are the masses going to get capital to buy the stakes in the first place if they are “not needed” at all.”

        In a democracy, all they have to do is vote for it – expand unemployment, then welfare and social security. With adequately increasing wealth produced mostly by machines, that should be more (politically) palatable than it would be today.

    • Jay


      Economically, that vision may be possible. The question is in how the benefits are distributed, which is a political question (arguably _the_ political question).

      One might imagine two different equilibria. One equilibrium (let’s call it “European”) involves distributing the benefits as widely as possible, probably as a result of many deliberate government policies. Another (let’s call it “Argentinean”) involves a small ownership class and a large class that struggles for subsistance. As my names indicate, both models can be observed in real life, and both are at least somewhat stable.

  • And once again, Ricardo’s Law of Comparative Advantage demonstrates itself the Pons Asinorum of macroeconomics.

    • Doug S.

      One constraint on the law of comparative advantage is that it only applies if your marginal productivity isn’t zero – you can actually produce something that other people want. Here’s an example where that doesn’t hold:

      I own and operate a factory that takes wheat and fuel as inputs, and turns them into bread. For every 5 units of wheat and 5 units of fuel, it produces four units of bread. Furthermore, my factory has effectively infinite capacity; I can turn wheat and fuel into bread faster than the local economy can produce wheat and fuel. However, I have no other skills; I literally can’t do anything of economic value other than turn 5 units of wheat and 5 units of fuel into 4 units of bread.

      Now, my neighbor goes and builds a factory that turns 5 units of wheat and 3 units of fuel into four units of bread. It, too, has effectively infinite capacity. I’m screwed, and Ricardo’s Law can’t help me. My marginal productivity has just become zero, because I can’t produce bread unless someone gives me wheat and fuel, and giving wheat and fuel to my neighbor produces more bread than giving wheat and fuel to me, so nobody will give me wheat and fuel.

      • Suppose you’ve got a bunch of humans on a desert island, managing to raise enough food to get by. Nothing you can do from outside that desert island can make them all worse off unless (a) you declare that they no longer own the island (b) the people on the island do something that a homo economicus wouldn’t. Both A and B are very real possibilities, but you have to say what B is exactly – what the people are doing wrong.

        Some years ago, I worried that increasing productivity would result in a spiral of unemployment: Automation means 10% of the people lose their jobs, then decreased demand means that 9.1% of the remainder lose their jobs, and so on.

        And Peter Thiel said to me, “But that would violate Ricardo’s Law of Comparative Advantage.”

        The difference between myself and this Ford fellow is that I immediately understood what Thiel meant and that he was right, so I gave up and went away.

        A few weeks later I came back and said, “Peter, all unemployment violates Ricardo’s Law of Comparative Advantage.” So we may very much indeed be doing something wrong. And if we’re doing it wronger than ever before, that may explain the recent wave of recessions followed by jobless recoveries – in contrast to the earlier and successful and economically normative case of agricultural employment going from 98% to 2% with an attendant huge rise in standards of living. And if we go on doing that thing wrong, we could get in more trouble later on. But you have to at least acknowledge that it is neither the economically normative outcome nor the past historical outcome. If I’d stared blankly at Thiel I wouldn’t have blamed him for writing me off as economically illiterate.

      • Dan


        Hmm I thought it over a bit :).

        I can see scenario A happening. The stupid humans is wasting natural resources, we machines can use it better. So Doug S. scenario, people need resources to live they die. Your paperclip scenario essentially.

        But that is the case of of intelligent machines.

        Regarding Ford’s scenario, Comparative advantage and Creative destruction seems to be powerful rebuttals, yet when you look at the real world you see constant output gaps, jobless recoveries and structural unemployment(but lowish).
        The logic seems perfect yet Ford’s rebuttal to the logic of “How am I going to pay the mortgage” has a powerful resonance…

      • Eliezer, Ricardo’s Law, like other economic and social “laws”, describes how things work within the underlying human social-political framework. Change the underlying framework enough and the laws may no longer be valid. We need work that describes the underlying framework more precisely before we can really predict what changes and where will invalidate the “laws”.

      • Doug S.

        Indeed, you can’t make them all worse off, but you can certainly make some of them worse off.

        In terms of “all unemployment violates Ricardo’s Law”, well, we all know that the job market isn’t exactly frictionless. Someone who’s been automated out of a job might not be all that eager to start waiting tables instead of looking for something that pays better…

      • Doug S.

        Okay, I’ve figured out a way for the outside world to screw over a group of humans on an island, without the inhabitants doing anything “obviously” stupid.

        Outside world: You’ve got a lot of apple orchards on your island, but the world market price for cotton is a lot higher than for apples. If you cut down your apple trees, plant cotton, and sell it to us, you can buy use the money to buy both apples and cotton clothing from the rest of the world.

        Islanders: Sounds like a good deal. We’ll do it.

        So, all is well, for a while, until…

        Outside world: All our cotton clothing factories were destroyed in an earthquake, and we’ve decided that it doesn’t pay to rebuild them. We suggest that you go back to growing apples.

        Islanders: But apple trees take thirty years to grow, and we don’t have the technology to do anything useful with all the cotton that you don’t want any more.

        Outside world: That’s not our problem. Maybe you could feed yourself by growing corn?

        Islanders: Well, we could do that, but we like apples a lot more than we like corn, and our land isn’t very good for corn, and…

        Outside world: As we said, that’s not our problem.

        So the islanders end up much poorer than they were, while they wait for their apple trees to grow back. Maybe they shouldn’t have cut down their apple trees, but it doesn’t seem like a mistake homo economicus wouldn’t have made, since the island had a clear comparative advantage in the production of cotton…

      • Jay

        Doug, wishing to reply to your reply…

        I have a PhD, and I’ve been out of work a while. It’s not that I don’t want to flip burgers (my aversion to low-wage work is no greater that most people’s), it’s that I can’t get hired for low-wage jobs. The low-wage employers think (correctly) that I would leave them as soon as I got the chance. They don’t like turnover, so they hire people with fewer skills (and fewer opportunities if the economy turns up). You might even say that the unskilled have a comparative advantage in unskilled labor.

      • John Maxwell IV

        Jay: Can’t you wear clothes from Goodwill, learn some slang, and leave your degree off your resume?

      • Tracy W

        Well the theory of comparative advantage is a theory of comparative advantage, in other words both members have to be able to produce at least two things. The point of the theory is to address the situation where one party is more productive than the other at everything.

        The situation of you only able to produce wheat strikes me as more equivalent to the situation of supporting people who are incapable of supporting themselves. A more realistic comparator to your hypothetical person who can only produce wheat is, say, someone in the post-traumatic amnesia stage of a severe brain injury who can’t produce anything. The only answers I have heard for that situation are charity or government provision of a basic income.
        As a practical matter, I can’t think of a single case where a real human being can produce wheat but not anything else.

    • Stuart Armstrong

      The risk might be that the marginal productivity of humans declines below the amount needed to sustain life. We have certain repeated fixed costs that we can’t really compress, and we need to find ways of paying them.

      Someone once quoted the example of horses once the automobile arrived, and the precipitous population decline that followed.

  • kebko

    Why do you even address the issue of falling wages? The fact that wages for blacksmithing and ice delivery have fallen dramatically would have scared the snot out of Martin Ford’s great-grandpa, but it’s quite irrelevant to Martin Ford. Addressing his supposition that overall wages will fall is giving him way too much credit.

    • The issue is causality. Falling prices lead to falling wages. Thus, it takes a more adept argument to claim that real wages (amount of goods / services recieved per unit of labor) will fall in the fact of automaton / slave labor. Thus, while today’s factory worker is paid lower wages than the factor worker of the past (relative to other professions), the factory worker can purchase many more goods, because of the price of such goods has fallen.

  • Michael Turner

    When Ford (ironically named) writes of declining wages, does he mean a decline in terms of purchasing power? But wait a sec: lots of automation means more purchasing power, all other things being equal. Right?

    Someday I might be so automated out of any job I might be able to do, I’ll have to sell my feces for their fertilizer value and run on a treadmill a couple hours every morning to sell the energy to the local power company, just to be able earn enough to keep a (robotically-built roof) over my head, put (robotically-grown) food on the table. and still have enough money left over to buy (robotically-derived) raw materials for my hobby of blacksmithing Japanese swords, which I sell to my local chapter of the SCA.

    How dystopian. I think I’ll avoid the misery by killing myself now. (Where’s my Japanese sword?)

    • anon

      Yes, lots of automation increases GDP and purchasing power, but this doen’t help you if the value of land/natural resources spirals out of control due to them being the main input in robotized production.

      On the bright side, land/resource rents are easy to redistribute. So most humans can probably sit back and earn these rents while robots do all the work.

    • Dan

      When Ford (ironically named) writes of declining wages, does he mean a decline in terms of purchasing power? But wait a sec: lots of automation means more purchasing power, all other things being equal. Right?

      He writes about structural unemployment… if you earn a wage yes it will increase as your fellow workers are fired… that is until you are next in line… and it drops suddenly. Expelled from paradise if you will. Structural employment not declining wages, productivity etc…

  • Pingback: Why I think Robin Hanson is Wrong about “Economic Growth Given Machine Intelligence” « econfuture | Future Economics and Technology()

  • Pingback: The Economic Implications of Intelligent Machines « econfuture | Future Economics and Technology()

  • I just added to the post, responding to Ford’s response.

  • anon

    “Yes, a sudden unanticipated change would be disruptive, but no disruption does not imply falling production. Ford needs to learn some economics, or listen to some economists.”

    I think this reply is unclear. Most macroeconomists in a variety of traditions would agree that sudden disruption can cause production to fall below its long-run potential. On the other hand Michael Ford’s book seems to be about long run effects, which is a different matter.

  • Small comments, I don’t think have been explicitly stated, unless I’ve missed it from others..

    Robin Quotes Ford:

    “If wages fall dramatically, then consumption must likewise fall because the majority of personal consumption is supported by wage income..”

    This is only true in a world where the cost of consuming goods and services does not fall in tandem.

    If technology displaces workers and is able to produce things for 1/10th the cost, competition would force deflation and competition in the new labor market would drive me to gladly accept a new salary aprox1/10th what it used to be. I would still be able to consume the same amount as before and would be no worse off at the lower salary.

    Also it looks to me that the technological progress we are making, which is going to put deflationary pressure on both wages and the costs of goods, is eventually going to run up against our minimum wage laws if it hasn’t already. We could run into a situation where not only does technological progress displace one set of workers directly, but then by pushing the equilibrium wage rate below our wage floor it would also exacerbate the unemployment problem by displacing another set of workers whose jobs are now too expensive to keep.

    • Jay

      More to the point, technology changes the mix of the basic factors of production in goods and services. In general, most businesses are aiming to reduce the amount of labor in pretty much every good and service, with much less attention paid to decreasing capital requirements or raw materials. It would be stupid of them not to; labor is expensive.

      It’s not crazy to think that, overall, this process will result in lasting disadvantage to sellers of labor (most people), with owners of capital and providers of raw materials capturing more of the value of the economy. The last time I checked, even China was losing manufacturing jobs; machines are cheaper than even Chinese labor.

      • Tim Fowler

        Jay – Re: “It’s not crazy to think that, overall, this process will result in lasting disadvantage to sellers of labor (most people), with owners of capital and providers of raw materials capturing more of the value of the economy.”

        Its certainly not crazy to think that it may result in a lasting or growing relative disadvantage to sellers of labor, but if production gets cheaper that relative disadvantage can still be with an absolute advantage as you get a smaller portion of a bigger pie, and your piece still is larger even if its grown only half, or even only a millionth as much as some others.

        Also people, esp. over the long term, respond to such changes and incentives in complex ways. Even just in terms of voluntary economics people will learn new ways to be useful to those with wealth. Also you have the possibility for political redistribution.

  • Bill

    It has been a recurring theme of economic history predicting that machines, and now computers, will eliminate employment.

    It might be useful to review this literature, and see what actually transpired.

    Futurists need to become historians who examine past futurists to make them more sceptical of their own predictions and to find out why past futurists missed the point.

    Here’s an example (and you can think of this as a Says law of intellectual progress, in that supply of new technology creates its own demand): did the personal computer, by bringing computing capacity to the masses, reduce demand for computation, or increase it; did it increase demand for more software, or reduce it?

  • Kling’s quote, “I agree that Singularians are far too optimistic about artificial intelligence. It is a variation of the “fatal conceit” problem.” Is actually almost 180 degrees from the reality of some Singularitarians. I am pessimistic about AI, I would prefer it not to come about but I see it as inevitable, therefore I must work to handle the problem as best I can. And it’s not a “fatal conceit” problem – it’s more like how the Austrians have been predicting depressions and collapse for decades and everyone has ignored them, because their timing was too “optimistic”. (And everyone is still ignoring them.)

  • Bill

    I would like to add another point to artificial intelligence: and that is the issue of what I would call Trust.

    The interesting question is: is a human being more likely to trust another human being or an intelligent system programmed by another human being.

    I probably have more trust in a computer generated five day weather forcast than the prognosticastion of the inassisted weatherman.

    But, how would I trust the opinion of a doctor versus an intelligent system the doctor could not explain when it came to a course of treatment? I trust the doctor to realize the ramifications of death, injury, and feelings of a person in a disabled state in assessing risk, but how does an intelligent system categorize and weigh those attributes, particularly when, in the end, it may be a human being who does the surgery, for example, knows his/her own limitations, despite what a computer model says.

  • Bill

    The other issue on artificial intelligence is that it did not do too well in managing portfolios in the last economic crisis. Given that the input of financial models may be based on the irrational conduct (and I am doubtful it is strictly predictably irrational), I am skeptical of artificial intelligence unless it can beat the market.

  • Robert Koslover

    I think the idea that we can run out of jobs due to automation is: (1) mistaken, and (2) doesn’t even matter, anyway. First of all, there is simply no limit to the nuttiness of “jobs” that can be created at whim. Consider, as just a few examples, the “jobs” of: fiction author, football player, actor, guitarist, blogger, comedian, professional bowler, interior decorator, personal trainer, psychotherapist, aroma therapist, etc. None of those “jobs” do anything in terms of actually generating food, clothing, housing, etc. I.e., they are all artificial jobs in some sense, which can exist only because our economy is so fantastically productive that there is room for such luxuries and or trivialities. Ultimately, once the robots take care of all the boring “real” jobs and thus provide for all our actual physical needs, we humans can indulge in whatever activities we want. Call them jobs, if you prefer. But they will be more like hobbies. Equivalently, you could call it “retirement.” My parents are retired and do not have jobs. Interestingly, they do not spend any of their time lamenting that they are “unemployed.” What would be so terrible about a world where all humans are “retired” and thus free to pursue whatever hobbies they wish, while surrounded by a land of plenty, as provided by smart self-maintaining robots? And in such a world, I have no doubt that Robin would still have this blog, and there would still be interesting discussions on it! Again, who really needs or wants a “real” job in such a world? The job of pony express rider is gone forever. Well, is anyone mourning over it? The same goes for every other “job” that a robot will eventually do instead of a human. Good riddance to all “jobs” and welcome to true freedom.

    • You’re right that it would be great for people to do what they enjoyed. Bob Black talks about that in “The Abolition of Work“. The problem comes if our society assumes that the only legitimate way to get income to live on, unless you have a lot of capital, is through income producing work. As “The Triple Revolution memorandum” said in 1964: “The continuance of the income-through-jobs link as the only major mechanism for distributing effective demand — for granting the right to consume — now acts as the main brake on the almost unlimited capacity of a cybernated productive system.”

      More on these ideas on this blog: “Beyond a Jobless Recovery“.

  • A dude

    the rents will most certainly be concentrated among the few. The rest of the population will then have to quickly find a way to be useful to these few in some court jester capacity, stroking their desires for status and entertainment. If they can’t, Ford has a point.

    GDP per capita of the surviving humans will be higher, total GDP not necessarily so.

    This rhymes with the link posted in this blog a few days ago about optimal status games not being concentrated. Same way, if court jester skills are concentrated, the rest of the population fades away.

    • anon

      Yes, some wages/rents will no doubt accrue to people with rare abilities. But natural resource rent should also be a big part of the economy in a fully-robotized world, and political considerations will most likely require this rent to be broadly distributed.

  • A dude

    and yes, US is already there, with useful stuff being mostly produced by global slaves and automation. So far plenty of court jesters have jobs (85% of employment is in services).

    • It’s too bad that this blog doesn’t have up-voting, so I’m commenting to “up-vote” this. I was laughing and clapping as I read this. Yes, this seems precisely how it is.

  • From:
    “Why limited demand means joblessness”

    Summary: Mainstream economics assumes demand for almost anything is infinite. Thus, the theory goes, when human workers get replaced by robots, or better design means less human labor is needed, then there will soon be new jobs making new things; the only issue might be retraining. But, if demand is limited (because the best things in life are free or cheap, and everything you own also owns you), then when people get laid off, the jobs are gone for good, because there is nothing more that anybody wants then is already produced. And people having more time outside of compulsory work would be a good thing, if we more evenly shared the wealth from automation and better design, but we don’t — yet.

    These are some ways to deal with increasing joblessness…:

    temporary measures like unemployment insurance and retraining funds, and when those fail, letting people live with relatives who still have jobs or be homeless (the USA now has one million homeless schoolchildren, an amount that has doubled in the last two years);
    government public works like in the 1930s (infrastructure, arts, research, medicine, etc.);
    a “basic income” for everyone, essentially Social Security and Medicaid for all with no means testing;

    improved local subsistence like with 3D printing and organic gardening;
    a p2p gift economy (like Wikipedia and Debian GNU/Linux);
    a shorter work week (like tried in France);
    rethinking work to be more fun so it is done as play;

    alternative currencies or other forms of exchange like barter or more formal rationing;
    increasing advertising to entice people into more debt (one cause of the current economic crisis as the debt bubble burst);
    intentionally producing shoddy merchandise or things with planned obsolescence, perhaps encouraged by promoting faddism in the culture;
    more prisons (employs guards and keeps people out of the labor pool);
    more schooling (employs guards/teachers and keeps people out of the labor pool) while suppressing true education; and
    more war (employs guards/soldiers, blows up and wastes abundance, and kills or disables workers to keep them out of the labor pool).

    Likely we will see a mix of all those in the future, and in fact, a mix of all those is what we have now (not that the last five options of advertising, faddism, schooling, prison, and war are recommended, even as our society currently relies on them heavily to destroy abundance and create guarding jobs). …

  • Pingback: Tweets that mention Overcoming Bias : Take Both Econ, Techies Seriously -- Topsy.com()

  • While I agree with you that Ford’s intuitive approach to economics leaves much to be desired I think you could be more charitable about interpreting the argument he is making.

    I would take him to be saying something like the following.

    1) Technologically advanced products tends to have large fixed costs, e.g., microchips require multi-billion dollar investments in fabs.

    2) The very wealthy devote a much smaller share of their income to purchasing the outputs of any particular high tech process than do those of moderate income. In other words the wealthy don’t buy many more microchips/cell phones/etc but instead engage in other kinds of consumption (sail boats, expensive dinners etc..).

    3) Therefore the more wealth becomes concentrated the less technological gizmos will be produced. He is probably actually claiming that the percent of income devoted to capital will actually fall. Whether this is a ‘failure’ or simply a shift towards an economy that produces leisure craft rather than iphones is a matter of perspective.

    I’m not convinced by the argument but it at least seems none ridiculous when phrased this way.

    Note, that when he says wages fall dramatically for the vast majority of workers he’s not saying the proportion of income devoted to labour is decreasing. It’s just that the distribution of wages has changed.

    • Hmm, maybe I was being too charitable.

      After all he can’t be both worried that robots will take all our jobs and that investment in capital intensive high tech processes will fall precipitously. Also I think that if you work out the above worry quantitatively it would turn out not to be a problem (self-correcting).

    • anon

      Yes some products have large economies of scale and as the economy becomes more diverse these products will face shrinking markets. But this has little to do with increased inequality: in a very unequal society the poor may well account for a large fraction of the economy.

  • Also I’m not so sure economists should listen to techies. After all Moore’s law and similar ‘dumb’ estimates of future performance seem to consistently do as well or better than educated guesses by techies.

    Moreover, economics is a academic discipline with theoretical underpinnings and often an uncontroversial received view. The same can’t be said for ‘tech.’ While there are various formalisms for computation, academic fields about circuit design etc.., there is no established academic discipline for predicting broad trends (will computers be able to replace most human jobs). Certainly one wouldn’t want to assume that having a degree in AI rendered your predictions particularly reliable, quite possibly the selection bias does exactly the opposite.

    So while techies who ignore economists are offering up their intuitions instead of the predictions of a well developed successful discipline economists who ignore techies are simply substituting their guess for the guess of someone who works with computers. Frankly, I’m more inclined to trust Tyler Cowen’s intuition on the future of computers in the workforce than most programmers or CS professors.

    In fact whether or not robots will outcompete people isn’t even really a tech question. It’s as much about sociology and public policy as it is about technological capabilities. I mean it depends on whether we choose to augment our own capabilities and how accepting we are of robots performing service jobs.

    • In fact whether or not robots will outcompete people isn’t even really a tech question. It’s as much about sociology and public policy as it is about technological capabilities. I mean it depends on whether we choose to augment our own capabilities and how accepting we are of robots performing service jobs.

      A common misconception by those with little understanding of AI.

    • Cowen doesn’t explain himself. His view seems hard to defend to me – I wouldn’t put much stock in it.

  • Steven Hales

    Don’t we have an example of automation from the goods sector. The labor content of all goods continues to fall across the world yet employment in the service sector continues to grow. It is easier than ever to assemble resources into finished goods. The labor inputs to manufacturing are now dominated by a thin layer of management, R&D, marketing and financial risk management. If AI begins to replace any of these jobs then that AI must “feel” in someway how a customer will perceive a new product. Usefulness is not enough. Tactile experience and elaborate feedbacks that characterize human life and experience must be present as well. A technocractic approach to new product development will never work it will produce “gray” products for the blade runner world. Overall creating such a copy of a human seems awfully rube goldbergish to me when we have a perfectly good replication system today.

    What if just good enough AI and Robotics begins to replace service sector jobs where will the new jobs be? Are we headed to a new classification of “service” where human experience dominates? If AI can never experience humaness though it could be conscious and “smarter” what is “smarter?” Is it sheer mental acuity the ability to solve physics problems? Will we suddenly have a simpler solution to Fermat’s last equation, a unified theory of everything? But without the human capacity for wonder and joy such solutions are without content they can and do exist without us.

  • Steven Hales

    Another question that always nags at me is that if we successfully create a human in silicon that feels, is conscious, and is smarter than we are why should it work for us? It seems to me that it would look at us as we look at gorillas. It might grow food and provide for certain necessities for a viable breeding program but what would motivate it to do anything beyond that? Wouldn’t it favor the inorganic over the organic. What preferences would an AI have? What are its indifference curves?

  • ChrisA

    To take what Ford is worried about to its logical conclusion imagine that there exists a machine that can produce any good or service requested of it (food, fashion, books written in style of favorite author, economics research etc) at essentially zero input, i.e. for free. If everyone had a machine like this there would be no employment at all (because any human business would be outcompeted by the machine) but of course it would not matter, since everyone could get whatever they wanted from the machine. Only if you believe there is some spiritual value to employment would this be a problem.

    Now imagine that these machines are only owned by 10% of the population, the remainder are economically barred somehow from owning one (they could not afford it perhaps), and have to survive with normal technology. So Ford’s argument goes, the remaining 90% would all be outcompeted by the 10% and they all would be unemployed. But they can only be outcompeted if the 10% gave all everything away for free, in which case we are back in the first situation, i.e. everyone is unemployed, but that’s fine because no-one needs to work. If the 10% decided to trade with some of the 90% (i.e. don’t give away their stuff for free) then at least some of the 90% have to be employed as otherwise they wouldn’t have anything to trade (hard to see why the 10% would trade with the 90% rather than give stuff away because in theory they already have by definition anything they could want, but let’s go with it anyway). Let’s say the rich 10% arbitrarily set their prices so that 90% of the remaining 90% can afford their goods (we must also assume of course that the 10% are a cartel otherwise they will compete each other down to free again). This is exactly the same as saying that at least 81% of the population must either be in paid employment (or have enough investment income or be a dependent) to be afford this price. So no problem for at least 81% they have all the goods and services as the 10% (except that they are of course working when perhaps they don’t want to, but it solves our spiritual problem if it existed). What about the remaining 9%, do they starve? Of course not, they have our technologies and we don’t starve, we trade among ourselves and do OK, so they are employed as well although they likely have a lower standard of living. Obviously if you vary the percentages the point remains the same.

    In fact if you look at the world today we see this same issue. We in the west have access to almost magical technologies to produce things that someone in a poor third world environment could not compete with on cost. One example is weaving. A hand weaver cannot possibly compete with a power loom you would think. Except that people in many places are still using hand looms, because of exactly the same processes as described above. The hand weaver is in a different economy which is largely not trading with the power loom economy.

    Of course the rich 10% in this example have tremendous power compared with the rest, but this is a political not economic issue. As noted by others above we face this issue today and democracy, with varying degrees of success, seems to cope. I would suggest that the reason we still have these global disparities such as the hand weaver example is due to the fact we don’t have a world democratic government to arrange transfers, since we don’t usually see this huge disparity within countries.

    • SmokeyTBear

      That’s more or less along the lines of my analysis on the matter, but I find myself wondering, if that is the case, why do countries like Spain an Greece have nearly 50% unemployment for young people? How is it that with this many people, a “sub-market” as you’ve described doesn’t produce itself?

  • consider

    In fact whether or not robots will outcompete people isn’t even really a tech question. It’s as much about sociology and public policy as it is about technological capabilities. I mean it depends on whether we choose to augment our own capabilities and how accepting we are of robots performing service jobs.

    A common misconception by those with little understanding of AI.


    • Doug S.

      Short answer:

      Strong AI : humans :: humans : chimpanzees.

  • consider

    (I should have added, he wasn’t necessarily talking about AI, but much more productive robots/devices.)

    • It sounds as though you are talking about relinquishment.

      Relinquishment is an option on paper, but those that try it tend to get marginalised and out-competed by those that embrace technology. Maybe a powerful world government could enforce it – but we would need one of them first – and large governments are not known for their rejection of technology.

  • I liked most of the Ford material I found. It is fairly simple stuff, though – and he seems to make one substantial mistake – that automation will put consumers out of jobs – and so will destroy the world’s markets.

    But what about machine consumers? It’s as though Ford hasn’t even considered that machines will buy things as well as doing the world’s work. Markets for energy, materials, spare parts, engineers, etc. seem likely to boom.

  • Steven Hales

    Is all of this about the end of scarcity? Is it the end of the economic problem? It is assumed that human wants are infinite and satisfactions are constrained by relative scarcity. Is the age of intelligent machines, nano factories and genetic repair to prevent senescence really the end of the economic problem? What will be the value of anything if costs to produce them falls and continues to fall? If we look at it as Ford does as a sudden change where just “good enough” tech supplants work then we have mass disruptions. However, if we look at it as a continuum then what investments will be made if his scenario of falling consumer demand is true? The feedback loop of consumer demand and investment is strong and because of this Ford’s scenario can never happen. But what will happen? We expend much less effort in time and money today to provide for the basic necessities of life so in some sense much work has already been supplanted by machines. But because human wants are nearly infinite we are willing to work to buy other things like a gourmet meal at a fine restaurant or the latest tech gadget. We are willing to work to stream more services into our homes from broadband lines to satellite TV. We are also willing to retrain to perform new jobs as the number of people looking for employment fell recently because some are enrolling in a variety of educational programs. All of this is driven by our nearly infinite wants. Ford’s problem is that he ignores the economic problem. If there is no work then the reasons for no work will never materialize because the investment to create the condition for no work will never materialize. Instead of a feedback loop Ford has broken the circle and created a finite line. What Ford is looking at is incremental change and he is extrapolating a condition that can have no cause. It is a sterile analysis. In Ford’s future investment will have to stop. Innovation will stop. The civilization withers and dies of its own lack of inertia. Ford in effect has repealed the economic problem by making scarcity absolute in the face of unlimitted abundance. In Ford’s strange universe effects emerge without cause.

    • Steven Hales

      oops, change “lack of inertia” to inertia.

    • Jay

      I wouldn’t think of it as “the end of scarcity” but as “the end of scarce labor”. Raw materials and capital would still have value, but labor would have almost none (not enough to manage subsistance, anyway).

      Many of the more populated parts of the world are already like this to some extent. In that type of economy, the initial resource distribution (and ongoing redistributions) matter enormously, because “working your way up” is all but impossible.

      • Steven Hales


        You cannot bootstrap the future without incentives for investment.

        I don’t think you get the point of my argument. If you accept the tight relationship of forecasts of future demand and current investment then if forecasts show a drop in demand from permanent unemployment then the investment won’t be made in the first place. The whole ‘end of work’ argument collapses in on itself. It is illogical. It also flies in the face of our experience through history. If the ‘end of work’ argument were true as stated then we would become Luddites as we would smash the technology and kill the robots. It would be an act of self-preservation. What happens is that as firms compete with each other, prices fall just as they did for textile manufacture when the automated loom was widely adopted in 18th century England. There the Luddites were responding to falling prices not automation but they were all former textile workers. In the ‘end of work’ scenario, production has to fall. If just good-enough AI allows a firm’s labor pool to shrink to almost zero then human labor no longer determines how much output can be profitably produced and shifts to how much AI can be added until its costs, rather than human labor’s wage rate, equal the firm’s marginal product. But the firm still faces the same market and must adjust production and this determines additional investment in AI. But in this scenario all firms have the same endowments and have no comparative advantage because just good-enough AI is driving all of the firms. If this is ubiquitous then all production stops because nothing is profitable anymore which gets us back to whether the ‘end of work’ scenario will never happen.

        But what is the directionality of cultural and economic evolution? Just as biological evolution does not move backwards neither does cultural or economic evolution. As in Bronowski’s barbed arrow of time, the barb prevents backsliding, we ratchet forward in a process called stratified stability. Application of these ideas to economics is not new; Schumpeter’s creative destruction and Veblen’s ideas are context specific in what drives economic evolution. In our lifetime we have seen the big established firms in almost every industry be toppled by nimble smaller firms. What drives that destruction but innovation? A large firm might try to adopt the new technology but it has a customer base mismatch that doesn’t support investment so the decisions made by management might seem rational and profit maximizing but they are not. The disposable digital camera is one example where it will threaten the established firms selling the newly designated non-disposable digital cameras. The established GPS firms like Garmin are currently threatened by ubiquitous GPS tech on mobile devices. The list goes on and on. When computer technology was entering the office environment we didn’t see an increase in output but when networking was added we saw an explosion in productivity, one technology leveraging another. Therefore, we have two complimentary processes in economic evolution, creative destruction and complimentary leveraging. How might a superior technology that could supplant all or most of human labor evolve? It certainly wouldn’t evolve through some kind of sudden scenario that the ‘end of work’ crowd thinks. To be successful and not discourage further investment it would have to evolve in a distributed process where all humans become the owners. Just as you and I own our own skill and its corresponding earning potential, we would begin to own this new technology. We are already purchasing robots of all kinds and gadgets of all kinds that have degrees of usefulness in performing human tasks. Don’t you think that the gradual distribution of technological capability will continue where each of us will become the owners of increasingly powerful technology? How does just good enough AI fit in to this scenario? I would say one it is first a niche market say in factory automation and the emergent flexible factory. At first it is expensive but the price point falls. It also simultaneously finds its way into decision support from finance to healthcare. It also finds its way into the home and everyday life. It immediately leverages both computer and networking technology but just good-enough AI cannot leverage its real abilities without advanced robotics, it must be able to move about. Once it reaches this level then ownership of the means of production becomes universal. What new jobs will there be for humans? Your guess is as good as mine is. However, to assume that there will be mass unemployment from a technological innovation is to fall back on an argument that productivity improvements will not free up purchasing power for other consumption. Those improvements increase aggregate demand and encourage more investment.

        If you believe in the human world of flesh and eventual mortality of family and community of cooperation of joy of laughter of sorrow of regret of conscience of reading of eating of tasting and savoring of fishing of sailing of exploring of children of birth of belief then you will see that it is not possible to reduce all of human experience to information because without the elaborate feedbacks of our physicality we would not have the sense of who we are we would not have that specific information in the first place. If we were to give that up or believed we could evolution would end.

      • Jay


        Of course most people have access to cheap technology. Still, very few people manage to produce much out of it, and the types of products that can be made with common consumer technology tend not to be worth all that much. There are many bloggers, many fewer bloggers worth reading, and almost nobody who makes a living at blogging.

        The few who have access to large-scale capital will, of course, resist being pulled down into the common fray with every tool at their disposal (think RIAA, think MicroSoft), with greater or lesser success.

        My understanding of a “cheap-labor” economy looks much like India, Argentina, or Russia. It is true that such societies have little effective demand, and that therefore there is little incentive to invest (excluding foreign markets from consideration). It would be inaccurate to say that “all production stops”, but economic growth is slow and demand is heavily skewed toward providing for the rich. I don’t see anything unrealistic about this model.

        I enjoy the physical as much as the next guy. The question is not whether machines can feel as we do, but whether they can work as we do (face it, your boss does not care much about your feelings, as long as they don’t get in the way of your work). For more and more people, the answer seems to be Yes.

        Evolution, properly understood, is far more prone to producing cockroaches than supermen.

  • Steven Hales

    In some of this discussion the transmission of problems, like mass unemployment, arises as a monetary phenomenon. If Ford’s scenario is correct then the velocity of money will fall and we will experience massive deflation. IOW the purchasing power per unit of accouunt will rise and the prices of goods and services will fall. This is a viscious cycle that would freeze decision making at all levels of society. In such a scenario the rise of intelligent machines would grind to a halt. The robot next door would be carrying a sign around saying “will work for energy.”

  • Dallas

    Remember, innovation is always slower in implementation that us technologist expect. In the case of information and software, the change can be rapid — there are no physical assets involved. In the case of intelligent automation of jobs, there are physical assets involved with far more material often devoted to his effectors (arms, sensors, tools, etc.) that to his “brain” or control system. These assets are neither free nor can be instantly produced.

    For pure information type jobs, dealing with a speaking machine won’t be any different that dealing with a bureaucrat — no compassion, reason or intelligence required. The rate of spread of this class of automation will be slowed by the power function nature of answers to questions as the complexity increases (they will take over very dumb jobs, but people in India will still be cheaper for more complex questions like why my software is crashing). This shift will also be gradual and not cause more than minor adjustments while increasing the total human wealth.

    Therefore, the rate of change will continue as in the past. Functions will slowly be changed from human to automated systems, just like toasters replaced holding a piece of bread over a stove. All the more steady state (invalid in the long run) assumptions of economics about market penetration, utility optimization, etc. will remain true.

    A bigger challenge than automation will be the increase in human population VS a finite planet.

    • Steven Hales

      “A bigger challenge than automation will be the increase in human population VS a finite planet.”

      Every new idea makes the concept of finiteness invalid. A single idea from one individual created the green revolution that continues to feed the world. More brains means a better world. Demographic armageddon is approaching for much of old europe it is approaching for Russia and China. Low fertility rates mean less future prosperity.

    • Jethro West

      ” dealing with a speaking machine won’t be any different that dealing with a bureaucrat — no compassion, reason or intelligence required.”

      This gave the first good laugh of the day!

  • Ricardo Law of Comparative Advantage start not working if taxation and regulation rise the costs of hired workers enough.
    When the cost to do something by themselves become lower than the cost to hire someone else and pay taxes.
    This is because there is a so high unemplyement in the western world.

    • Jay

      Minimum wage laws and regulations can cause unemployment, but are not the sole cause of unemployment. The requirements of subsistence set an effective minimum wage.

      In Victorian England, there was a widely understood principle that the equilibrium wage of the working class was the wage that kept the working population constant (the “Iron Law of Wages”). If pay goes down from that level, malnutrition and disease kills workers. The labor supply therefore falls and wages rise. The reverse takes more time, as children take time to grow.

      So far, this insight has not seemed relevant to North America, because the continent has been relatively unpopulated. Of course, the population keeps growing.

  • I thin Robin ignores three things:

    1. Utility curves. We know that above (approx) $15K a year, increases in income yield little increase in happiness. So those “luxury” goods are less valuable, meaning producers can’t sell them for enough (in aggregate) to make a living.

    2. The remarkable efficiency of technology-driven corporate capitalism is also a remarkably efficient pump pushing wealth to the top. With high enough productivity (one person with a computer making everything?), this could overwhelm the price cuts that people down the ladder experience.

    3. The massive productivity gains of recent centuries have been accompanied by massive increases in government redistribution. There’s not one thriving, prosperous country that doesn’t have a mongo huge dose of it.

    Is it possible that that redistribution is necessary to maintain aggregate demand so a high-productivity economy can thrive and prosper?

    Why Prosperity Requires a Welfare State

    • atari_eric

      Either redistribution is necessary, or (at the risk of sounding communist) the people need to directly own (or receive a share of the proceeds from) automation.

      • My vote for the second disjunct (sans parenthetical qualification).

  • Robotic Nation by Marshall Brain
    is a fun read and sounds very similar: Rapid progress in AI and robotics may bring about mass unemployment and economic collapse. It is well written, compelling, an thought provoking, and Brain gets many things right until he starts delving into the economics and his arguments seems to go astray.

  • Pingback: Asymptosis » Robin Hanson’s Reply to the Luddites()

  • Pingback: Overcoming Bias : My CQ Researcher OpEd()

  • Pingback: Martin Ford Asks: Will Automation Lead to Economic Collapse? « hims47()

  • Pingback: Welcome, Job-Destroying Robots | Messy Matters()

  • Pingback: Martin Ford Asks: Will Automation Lead to Economic Collapse?()