The Automation Myth | Hacker News Reader

The Automation Myth

vox.com

57 pointsvermontdevil47 comments

This article was very confusing and difficult to read. There seems to be a conflation of workers being replaced 'robots' (Human teller -> ATM) and the process by which humans are simply more productive through augmentation (Manager with paper ledger -> Manager with inventory system).

There are some nasty potential error sources when comparing productivity per country over time because they use different currencies with their own value changes over time. (Probably related to productivity, but lots of other factors)

And my biggest issue was the Wells Fargo twitter chart. Average growth per year is an ok tool, but it's simplistic. It gets really messed up when you consider a 5 year time frame 98-03 and a 10 year 04-14. With the straight averaging method compounded growth/loss gets messed up.

It's all the same thing from a high enough level. If your manager can do the work of two managers because he has an inventory system, you don't suddenly manage twice as much inventory (because everything else about your business doesn't change at the same time), you fire every other manager.

But the point is that society as a whole is richer because you can get a good from a manufacturer to a customer with less overall spending, and so while an individual inventory manager may or may not be able to get another job -- and might legitimately suffer -- ultimately the business does grow, and there are more jobs created than destroyed (though the created jobs may not be inventory manager jobs, and the displaced inventory managers may or may not ultimately benefit).

"and there are more jobs created than destroyed"

Why? Well, I know its tradition to say it, but there appears to be no causal relationship, just occasional coincidence. The mathematical model would be interesting to see.

Just occasional coincidence? Looks to me more like there's very common (almost uniform) coincidence, with occasional exceptions. But we don't actually call that "coincidence", we call it "normal behavior".

> you fire every other manager.

Except managers are "sticky", and generally aren't fired merely because they are less burdened. All jobs not just managers are in truth sticky, but in general the more status a job has, and the more high status interactions a job has the greater the difficulty in removing the position.

A lot of companies became much more productive on paper during the recession, not because they did massive changes, but because they were forced to finally cut some staff. Like a cheesy 80s movie they had the productivity inside themselves all along, but it took the need to trim budgets to expose that productivity.

Of course a lot of companies never actually really pared down things, or not enough to really boost their productivity. It just isn't simple to get that productivity to reveal itself, especially outside of an existential threat.

It's nice of worthless articles when they lie within the first few sentences to make it clear they should be disregarded.

> President Obama has warned that ATMs and airport check-in kiosks are contributing to high unemployment.

No, in the youtube segment linked he merely uses those as examples of jobs disappearing and thus creating an impetus for the government to stay on its toes about changes.

I think my explanation was better than Mr. Yglesias's, if I do say so myself: https://devinhelton.com/2013/01/20/economic-classes/

Automation has increased productivity in many areas. But there are multiple factors of production: labor, natural resources, zoned residential land near jobs. Plus there are other limits to production, such as monopolies and regulation. So if just automation improves, production only increases and prices only fall to a point, until natural resources, marketing, and monopoly profits make up the bulk of the cost. As a 22-year-old entering the job market, all you have is your labor to trade, but you need access to natural resources and constrained goods like housing. Since automation makes your labor less valuable, you actually need to work just as many hours or more to get the same purchasing power. Worse, you end up in a zero-sum game of marketing and sales, where everyone works just to get ahead of everyone else.

An underlying premise of productivity tracking is that we are able to consume more indefinitely. But in all likelihood there is a logistic curve to human material needs as well.

If we did have $30,000 more on average, as the article states, how would it be used? The case is obvious for the poor, but as you go to higher income brackets more and more of the possibilities fall into "luxury" and "status symbol".

Take, for example, the rise in the cost of college education. It's absolutely bizarre that in an era of low fiction information, tuition has actually gone up. That is, until you consider the credentialing as a form of signalling for limited jobs at the high end. The colleges aren't getting funded to expand the core mission of education, they're chasing each other to provide a premium experience that will attract people with a pedigree, who subsequently raise the status of that institution.

From that angle, automation's effect is to shuffle around the job landscape, not to directly increase productivity. More minds on higher value jobs - but eventually we start cutting into the highest value stuff we can think of. What automation doesn't do for us is expand our ability to be creative about what work is and what jobs could be done that aren't. That capability is directed through our social structures and "what people will pay for." Anyone good at conversation knows that you can have the same room and the same people and achieve wildly varying outcomes in discussion. You can have an economy that "shrinks" because less is measurably produced, yet people feel wealthier on average. Indeed, that's central to discussion of open source software and its commodifying effect.

In conclusion, no, we don't know. Social science is a fragile thing and the things that seemed obvious to one generation have a habit of being discarded by the next. Maybe the $9000 of income inequality is, in fact, the important number, even if we can fabricate another bigger number with linear extrapolation.

Who cares if it's a luxury good as long as it's not a positional good? If a larger group of people take, I don't know, out-of-town vacations or buy cool-but-not-very-useful electronic gadgets or take classes in glassblowing or firebreathing, then that's economic activity which in turn provides jobs for hotel staff, factory workers, and craftspeople.

Your single example is a university education, which is at least arguably a positional good. But tons of luxury items aren't positional.

> in all likelihood there is a logistic curve to human material needs as well.

Material needs, probably yes. But not all needs are material needs.

> as you go to higher income brackets more and more of the possibilities fall into "luxury" and "status symbol".

Those are loaded words; one person's "luxury" is another person's preferred entertainment.

> It's absolutely bizarre that in an era of low fiction information, tuition has actually gone up. That is, until you consider the credentialing as a form of signalling for limited jobs at the high end.

That's probably a contributing factor, but it can't be the primary reason tuition has gone up so much, precisely because, as you say, there are only a limited number of jobs for which this signalling is useful.

The primary reason tuition has gone up so much is the fact that student loans and grants are so widely available. Colleges have simply raised their prices in order to consume all that extra money. Most of those loans and grants don't go to people who view a degree as a status symbol or a signalling device; they go to people who, rightly or wrongly, sincerely believe that a degree will give them a chance at a better life, and the wide availability of financial aid means that more people are likely to come to that conclusion. (IMO, we've pushed that lever too far; many of the people now using loans and grants to get degrees will not get enough of a better life to make it worth it to them. But that's a whole other discussion.)

Status based economies are fragile.

There have been a few waves such as the Beatniks where people say eff status I want to relax. Electronics pushed that off for a while, but that treadmill seems to have slowed down. So, I expect the next wave of young people to accept they can live well on a 2 day a week job. https://en.wikipedia.org/wiki/Beatnik

Such is fine for 20-30 year olds without health issues or kids, but does not keep widget factories staffed or pay for pension checks.

This is something that I've always found odd about how the actual value in the economy is tracked. Specifically, how the stock markets continue to rise despite the decline in productivity. If I was told as a shareholder or private owner of a company, that an hour of labor and inputs are making me mess than I was making 20 years ago I would be upset but that doesn't seem to be reflected in the volumes and prices of the stock markets themselves. Why is this case? Is there something I'm missing?

All in all, I hope someone or some firms figure out how to break through the hurdle to automation in industries where it would matter like healthcare. It would be the first step to correct the economy in my opinion.

> Why is this case? Is there something I'm missing?

It's still always going to be better to have the portion of retail worker's production that goes to profits and shareholders, instead of the portion that goes to the worker, if for no other reason than that you can scale that side.

> This is something that I've always found odd about how the actual value in the economy is tracked. Specifically, how the stock markets continue to rise despite the decline in productivity.

I'm confused. Are you talking about the economy or the stock market? I don't see why the two should be related. Productivity is (unit of output / unit of input). The value of the stock market is the sum of the discounted cash flows from the market's member firms.

EDIT: marginal investor's estimation of

> Why is this case? Is there something I'm missing?

Labor is a market. It's segmented geographically and by industry, but you can think of it as a single market.

You might be 500% percent as productive as a person with your same abilities was 20 years ago, but so is everyone else. So you're contributing five times as much to your employer's bottom line, but since he's in a demand-constrained market (post-scarcity, pretty much everyone is), he doesn't need 500% percent productivity.

So he hires fewer workers. So the demand for workers goes down. So your skills are worth less.

This trend is going to continue until you have (hyperbole) a single person hitting a single button periodically as the sole employed person in the market. What happens to everyone else? No one knows.

> (hyperbole) a single person hitting a single button periodically as the sole employed person in the market. What happens to everyone else? No one knows.

Well, the owners would be the ones calling the shots to the intelligence systems.

I would gather that the country would look similar as Marshall Brain's Manna. Although, I'd take out the nicer parts about the "Australia Project".

Specifically, how the stock markets continue to rise despite the decline in productivity.

The stock market rises in nominal dollars, which is meaningless, since that is just a result of monetary inflation. Corporate profits, and personal income invested in the stock market will rise with total nominal national income. In economics terms, nominal national income growth is proportional to the growth in the money supply (M*V=Y). So the growth in the stock market is just a result of the growth of the money supply. It has zippo to do with any real thing like productivity. That is why it is always silly when newscasters report the stock market changes every day, and view it as good when it rises, and bad when it falls. They are essentially cheering on monetary dilution. (But inflation always feels good in the short term, so they aren't entirely wrong to cheer it on, it feels good when the stock market rises, even if the same underlying mechanic is also pushing up your milk price).

Thanks for clearing that up. It's one area where I can't seem to get a solid grasp on it (I know micro-economics and some macro-economics, but those don't seem to be helpful in understanding financing imo). So basically the stocks will rise regardless of productivity. But then where does productivity come into play? Maintaining the real cost of production below/at a certain threshold regardless of the nominal cost?

We've had very low inflation in the last 20 years or so. While certain numbers, especially short-term ones, are reported in nominal dollars instead of real ones (it's actually kind of hard to report on real dollars except in considerably retrospect), neither Yglesias nor anyone else in the broadest definition of economic literature is confused about the nominal/real distinction.

That said, productivity is not declining. The first derivative of productivity is declining -- that is, productivity is going up, but it's not going up as fast as it used to. Also, the population is going up. And that explains the (real, not nominal) stock market growth.

To make this clear: Productivity is not declining. The growth of productivity is slower than it was in the past, but it's still growing. The first graph here is annualized growth rates, and you'll note they're all positive: http://www.bls.gov/lpc/prodybar.htm

    The power of Moore's Law — which states that the power of computer chips doubles
    roughly every two years — is such that the next five years' worth of digital 
    progress will involve bigger leaps in raw processor power than the previous 
    five years. It's at least possible that we really will have a massive leap 
    forward in productivity someday soon that starts substantially reducing the 
    amount of human labor needed to drive the economy forward.

I wonder how long it's going to take for word of Moore's Law's end to filter out into the wider public perception. Here, in the industry, we look at Intel's delays going to first 14nm and 10nm as clear signs that Moore's Law is sputtering out (as all exponential growth does, eventually). But out in the wider public, there seems to be a perception that chips will get "twice as good" every year into perpetuity. It's going to be interesting to see when (or if) people realize that their hardware isn't getting better at the same rate, year-over-year, that it used to.

> But out in the wider public, there seems to be a perception that chips will get "twice as good" every year into perpetuity.

I'd think people in the wider public have no clue as to exactly how much faster computers get every year. Considering that software usually gets slower at about the same rate as processors get faster, I'd imagine many people would think the increases in speed are much smaller than they actually are.

Moore's Law might be ending, but what about the rise of heterogeneous hardware using multiple CPUs?

http://herbsutter.com/welcome-to-the-jungle/

That first graph stinks. It only counts job work hours worked per job worker, not work hours per person. In 1950 there were an awful lot of households with one adult working a paid job and one adult handling most of the household work. According to this graph, that one worker did 1900 hours per year plus a bit of household work, and the other adult did ??? hours of work. Now both adults are doing 1700 with the household work on top. Granted, the amount of household work has gone down, but has it really gone down by over 1500 hours per year? If not, we're working more, not less.

This is an awful article. Let me count the ways:

1. You don't measure productivity by the number of hours worked. There are just different cultural norms at play and it doesn't make sense.

2. Comparing an economy that started out with low productivity and grew quickly to an economy that started out with high productivity and grew more slowly isn't really telling me anything other than the low-productivity economy caught up. Which happens, because it's harder to push the frontier than follow in someone else's footsteps.

3. The effect of automation is very hard to quantify because it tends to displace low-skilled jobs that don't pay much in the first place. But the social impact of this is huge -- you take the people at the bottom of the economy and take away the only jobs that were accessible to them. Social unrest is inevitable. This will likely impact the economy in negative ways in years to come, and we need to understand this.

Anyway, all around a terrible article. The author has no grasp of basic economics and seems to have cherry picked statistics that he thinks back up his claims. Has anyone created a "top 10 logical fallacies of data analysis"? Because this article would tick all 10 of them.

> You don't measure productivity by the number of hours worked. There are just different cultural norms at play and it doesn't make sense.

The article doesn't. It measures productivity in terms of actual monetary value per worker and then compares it to the number of hours worked.

> The effect of automation is very hard to quantify because it tends to displace low-skilled jobs that don't pay much in the first place. But the social impact of this is huge -- you take the people at the bottom of the economy and take away the only jobs that were accessible to them. Social unrest is inevitable. This will likely impact the economy in negative ways in years to come, and we need to understand this.

Again, addressed by the article. Would you have considered milkmen or household servants high-skilled jobs? Both these jobs have been all but replaced for the majority of people, yet there wasn't that much social unrest when the refrigerator or washing machine was invented.

The author espouses this economic knowledge on his twitter feed:

"Never could have guessed from the name that the Communist Party of China wouldn't be all that interested in having a credible stock market."

That's all you need to know about reading this article.

This entire piece is relatively baseless. It's relatively trivial to say "If trends had continued in a straight line, things would be different, but they didn't, so they aren't."

Also, talking about average household income is a waste of time. Talking about average anything when you are talking about the working class is foolish. The average net worth of a household is five times the median.

The demand for automation slowed precisely because the demand for labor plummeted. That's the exact same thing as saying "The demand for oil plummeted because the cost of relative demand for natural gas plummeted."

Reading this article was a waste of time.

> Reading this article was a waste of time.

Agreed.

Productivity statistics are lagging indicators. Looking at them to predict the future is incomplete and short sighted.

The concern about automation lately is due to the R&D roadmaps of up and coming technologies. This is a leading indicator. Whether or not these new technologies actually catch on should be the debate.

He's talking about average rather than median precisely because his point isn't that "the economy grew, but the growth was captured by the wealthy," but rather "the economy hasn't grown much per capita." So you want to look at average, not median.

And he isn't saying that trends didn't continue, he's saying trends went one direction, but the policy discussion is as though the trends went in exactly the opposite direction, and that that's important.

This article encouraged my to look at why my life was so expensive.

1/3 of my expenses are related to living in Boston. Rent is 2/3 of my total expenses, and 1/2 of my rent is only because of the city I live in, if I chose a cheaper city or lived 15 minutes further away, I could have the same house for 1/2 the rent. Of the remaining 1/3 expenses, many of those are cost-of-living related as well.

Location is something that's never really going to get cheaper, because location is fundamentally scarce. You can have as many cities as you want, but you can only have 1 SF, 1 Boston, etc. I don't want to live in a city, I want to live in Boston. And that's expensive compared to other cities around the world.

I think this is a testament to how cheap technology has made our lives. 100 years ago, things like clothes and food were a much more significant part of lifetime expenses. And as technology continues to improve, location is going to continue to eat up a larger portion of total expenses.

Autonomous vehicles or high-speed subways that are cheap to build will destroy housing prices. Technology can reduce the cost of transportation increasing the usable land around a city.

Other advances include new construction techniques making it cheaper to build taller buildings. ie prefabs or 3D printing

Hyperloop! https://en.wikipedia.org/wiki/Hyperloop

Author has things reversed. Automation is affecting jobs with higher wages and higher productivity first, turning those markets into winner take all or eliminating labor altogether. The remaining workers are forced to crowd into lower productivity jobs and work longer hours for lower wages to make ends meet. Automation appears to cause lower overall productivity, but the actual effect will be quite obvious when the cost of automation is low enough that it comes for all the low wage workers.

Reply on news.ycombinator.com