"Automation crisis" reminds me of "peak oil." Take a current problem (high unemployment, high oil prices), ignore the reasons for them (the global financial crisis, disruption oil supply along with increased oil demand), and instead construct a strange eschatological reason for the problem (machines are taking all of our jobs, we're almost out of oil).
Let's look at actual measures of automation - productivity growth. It's been very low over the past decade[1] when compared to decades past. Prime age employment has been returning to pre-crisis highs[2] - despite people claiming that the old jobs were gone and new jobs wouldn't replace them (the machines were supposedly going to take them all).
I think you're oversimplifying and ignoring a few key factors here.
1) As has been shown time and time again, the rewards of productivity gains accrue to the owners of capital far more than to the workers who are more productive. The result is widening inequality on many levels, which extrapolated a few decades into the future ends up looking... not great. I hope we can agree that de facto serfdom is a bad thing.
2) You can't just look at the number of new jobs when making employment generalizations - you have to look at the quality and type of jobs that were created to replace the ones lost. It's like if you took away someone's house and then gave them a tiny apartment a year later, then claimed "Look, they got housing back! Everything's great again!". In the case of employment, the vast majority of jobs created in the last decade have been contract positions and low-end temp work [1]. If you look at the sectors where employment has grown when new jobs data comes out, low-end service jobs continuously top the list (ex. [2]). We can debate whether or not people actually want these types of new jobs as opposed to traditional real estate/financial services/salaried construction/etc. jobs that were lost in 2009-2010.
3) We haven't hit the automation crisis yet in any meaningful way. When autonomous trucks and warehouse robots become prevalent, then we will be in an actual crisis.
The automation crisis right now is not that everyone is going to lose their jobs overnight, but that automation is exacerbating big problems in the jobs market that lead to massive inequality, the prevalence of underemployment, and the trend towards bad jobs with few or no benefits and workers needing to work gigs on the side to make ends meet or afford the things that their normal job would have provided in decades past. When large swathes of workers lose their jobs to automation in the future, these problems are going to continue to compound unless we start making big policy changes. Yes, people will likely find new jobs to replace the ones lost, but will they actually be better or worse off than before?
1) and 2) seem to be largely effects of a globalized eceonomy, with outsourcing and resulting inequality. Automation seems to be truly a boogie monster you can put blame on.
Trains also took away long distance horse buggie jobs. They were largely automated too. And finally, if you think driverless cars are nearby, then so are completely humanoid robots who can do all the work a human can do.
I'm confused. What's the difference between outsourcing and automation in your mind? They are both driven by a desire to lower costs. Automation isn't yet "there" in terms of being able to implement a model/robot/system as effectively or cheaply as sending it to a lower labor cost country, but that's a temporary state of affairs. As soon as automating things is cheaper than paying low-cost labor, companies opt for automation. We've already seen this happen with manufacturing "coming back" to or staying in the US, when what actually comes back is highly automated factories with few blue collar jobs [1].
You seem to be arguing that the issues I brought up are tied only to globalization and outsourcing when they are, in fact, tied to underlying economics that strongly favors automation in the long run. Over the next 10-20 years as automation becomes cheaper and more practical, it makes sense that the same problems will accelerate, even as globalization and outsourcing decline.
Outsourcing means a country losing jobs to another country. Loss in one country, but not inevitable. It is just fair market competition.
Automation is global reduction of available jobs, which we might be doing (with agricultural mechanization mostly). But nowhere are we close to a mass change in society worldwide, not due to AI, and specially not to be given the blame of rising inequality, wages, debt, and unemployment.
Funny you mention horses, as it is a common analogy in these discourses. The horse population actually declined as they became less necessary thanks to transportation innovations. CGP Grey explained this well: https://youtu.be/7Pq-S557XQU?t=212
Everyone in the world used to be a farmer. Now, only 1% of the population in developed countries farm. This time it's different?
Actually, I can agree that there had been reduction in global jobs as world gets slowly automated. But there won't be a sudden takeover of all jobs by some super sentient AI leading to mass unemployment. And neither are the current problems of inequality, debt, low wages a result of an increasing productivity, but rather increasing globalization.
It's the exact same thing that is going to occur with the human population. The current popular calculations for global population growth and the reduction of that growth, are wildly wrong. The reduction in growth and eventual decline is going to happen sooner and will be far more aggressive. The horses were replaced, the humans will be replaced.
What I really want to read about is why otherwise intelligent people lazily default to pattern matching when changing facts suggest they should change their mind.
And I think the red zone is a century away. We have no understanding of how our brains work, yet alone making devices as smart as we are in causal reasoning.
That's the misconception. AGI is very unlikely to happen in our lifetimes due to prohibitive physical limits. But you don't need AGI to disenfranchise more or less 90+% of humans.
I find smarter people tend to broadly overestimate human potential. We are unreliable, tire, are often untrustworthy, make mistakes, are subject to emotions, prone to illogic and irrationality, etc. We're a very suboptimal entropy-reducing machine relative to the forms of purpose-built AI and robotics that will be possible within the next 20 years. It's a shame people aren't more alarmed about this.
> I hope we can agree that de facto serfdom is a bad thing.
There is actually some evidence that what we currently have is already worse than serfdom [0] in many ways. I don't think this is a crisis which is about to occur, but one which has already started, is currently ongoing, and will probably get much worse.
You bring up important issues, but there doesn't seem to be an reason to think they are the result of automation. There have been periods in the past with higher levels of productivity growth that have corresponded with a decrease in inequality. Now we're in a period of low productivity growth corresponding with an increase in inequality.
As for self driving trucks, I remember people telling me a decade ago that they would have lead to mass unemployment by now. The predictions of an impending automation crisis haven't had a great track record.
Is there a reason to think they are not the result of automation? Take a look at [1] - robots brought into US factories are directly responsible for lost jobs and lower wages. And I think you must be pointing to a state of affairs that hasn't been true for at least 40 years - look at the chart in [2]. I'm not interested in what may have been the case 60 years ago in post WWII boom-America, I'm interested in what the trends have been during the lifetime of the average reader here.
As for self driving trucks, who told you that? Per my prior comment, we're not there yet, but it is definitely coming. Automation hasn't reached driving yet and it won't in any significant way for at least another 5 years, but as soon as the tech is ready, adoption is going to be rapid. But saying "it hasn't happened yet, so clearly it's bogus" is like building a cabin on an active volcano and saying, "they said it would erupt within the last 5 years and that didn't happen, so clearly it's not going to erupt ever". Just because a predicted timeline is incorrect does not mean the predicted event won't come to pass.
A decade ago self driving trucks existed only in thoughtspace; today they're actually being built, driven, and refined. That the old estimates we're wrong doesn't mean the tech will never happen, it's almost available right now. Long haul trips are arguably the easiest to automate, since the routes are stable and easily mapped, even if the last few miles or exceptional conditions will still require human intervention (which could be remotely handled).
Betting on widespread automation of trucking is probably one of the safest AI-related bets you could make right now.
> 1) As has been shown time and time again, the rewards of productivity gains accrue to the owners of capital far more than to the workers who are more productive.
This looks much more like the real problem to me. The solution: Make it much easier and less bureaucratic for such people to invest their money.
That's not a solution unless they also decide to invest their money on terms better for the borrowers, which is not on the table, which is the actual problem. Nothing short of a philanthropic outburst by the worlds elites or some kind of truly major crisis will change the direction our trickle-up economics are sending us in. A more efficient model of investment is just a way to increase the effects of efficient investment, which is money going to the rich.
Unless you believe more people are wealthy enough to live fine without needing to work (the opposite seems to be true[1]), what matters is not the employment rate, but labor participation rate. That's not even considering whether average wages for those who are employed are are keeping pace with rising living costs. The good news is that, eventually, automation will reduce living costs. The bad news is that fewer people will have jobs enabling them to afford even those reduced living costs.
From the BLS website glossary: '"Productivity": A measure of economic efficiency that shows how effectively economic inputs are converted into goods and services.'
I don't see why productivity as an economic measure would increase in the early stages of a process where narrow AI and robotics start to take over human jobs. Until costs come down, when AI software and robotics becomes more generic, and standardized, and adopted outside of narrow industries, the economic effects won't be as dramatic.
The only escape from narrow AI and robots taking human jobs is the creation of more bullshit jobs (see: Graeber), or genuinely new job sectors, but I have yet to read any compelling predictions of new job sectors that will employ large numbers of people and can't be automated.
Creator jobs? How many people on youtube, or patreon, or self-publishing books, earn even minimum wage from the time they put in? If you dumped 1 million, or 10 million, additional people onto those platforms and told them their new job is to create full time, how many of those additional people could make a living wage, or even minimum wage, for the time they spend?
Service jobs? While the wealthy (let's say, upper middle class and above) might want human waiters, human voice support on the telephone, etc., I think that's vastly overplayed as an opportunity for human labor. I don't think that many wealthy people would prefer to deal with humans over good AI, and I don't think it would create that many service jobs even if wealthy people did mostly prefer to pay a premium for human service.
[1] With increasing numbers of people working beyond retirement age, more debt, and more couples both working (I have nothing against that, I'm just pointing out the labor statistical reality that families' standard of living requires that more, now).
> I believe this is the second graph you should have used: The labor force participation rate.
Here's labor force participation for prime age workers[1] (which is similar to the prime age Employment-Population Ratio I posted above). The problem with your graph is that it's including everyone 16 and over - so baby boomer retirement makes an impact, as does the ACA (college students not working a job for benefits because they're covered by their parents, people who are retiring earlier because they don't have to wait for Medicare).
Though it's worth noting that even if, for some reason, you wanted to look at everyone over 16, the participation rate is still well above anywhere it was in the 50's, 60's, and most of the 70's. If you're looking at prime age, we're only a couple of percents below the highs in the 90's (and much higher than other periods).
> but I have yet to read any compelling predictions of new job sectors that will employ large numbers of people and can't be automated.
I'm always confused by this argument. Practically every day on HN there are articles about problems that it would be great if more people were working on. Replication crisis? Aging infrastructure? Global warming? Cancer research? Environmental destruction? Food scarcity? Documentation and bug hunting in software? Taking care of frameworks? Early childhood education? Heck, even beyond all the problems on Earth, there are plenty of people who want to work on things like sending people to Mars.
I just can't fathom how people can look at the world and say, "Yep, we don't really need to do anything more than what we're already doing."
I think the trouble is the majority of those things are not new industries creating consumer goods or services. The majority of them are public goods that are hard to monetise. So they get neglected. Assuming those things can create more jobs assumes we that get better at allocating resources to them (either philanthropically, through the government or a societal way not yet dreamed of).
So while there is lots to do, it seems hard to think of huge numbers of jobs being created.
> Though it's worth noting that even if, for some reason, you wanted to look at everyone over 16, the participation rate is still well above anywhere it was in the 50's, 60's, and most of the 70's. If you're looking at prime age, we're only a couple of percents below the highs in the 90's (and much higher than other periods).
Because women entered the work force. Compare your linked chart to the same chart for women [0] and men [1]. It's an important dimension to the discussion, I think, since non-working women are generally socially respected, while non-working men are not.
> I'm always confused by this argument. Practically every day on HN there are articles about problems that it would be great if more people were working on. Replication crisis? Aging infrastructure? Global warming? Cancer research? Environmental destruction? Food scarcity? Documentation and bug hunting in software? Taking care of frameworks? Early childhood education? Heck, even beyond all the problems on Earth, there are plenty of people who want to work on things like sending people to Mars.
Almost everything you listed there requires high skill/education levels. What about service workers/low-skill jobs (y'know, the things that would be replaced by automaton)? I'm pretty sure they're the people GP was talking about.
>The good news is that, eventually, automation will reduce living costs.
I’m not sure even this is true because of the degree of hyper-financialization and rent-seeking in areas like housing and healthcare. For example, with housing, even assuming we clear the zoning barrier, you still run into the problem of housing being an investment vehicle for global Capital. They’ll let it sit vacant or be consolidated into rental conglomerates in exchange for a relatively secure asset with marginal returns.
It’s possible that systems in Europe with a greater degree of decommodification will benefit from automation (e.g. easier to build public housing and easier to administer socialized medicine), but this seems unlikely to hold true in the American context, barring some massive political change.
Leaving the US is already a great option for many people. So, the global response to automation is arguably more important than what any specific location does.
"Peak Oil" is a theoretical model for analyzing oil supply and production cost. It was first conceptualized in the 1950ies, in a time where hardly anyone was claiming that we're almost out of oil. Indeed "Peak Oil" does not claim that the absolute oil reserves are nearly empty but that the reserves that can be exploited cheaply are at their maximum production capacity. What this model does not take into account are price fluctuations, based on a decrease in growth and economic activity (what we have seen in the past 10 years) and new technologies (Fracking, LNG). So this model is imprecise and incomplete, but useful when analyzing trends over a very long period of time.
You claim, that the "Automation crisis" isn't a real problem either, because apparently employment is almost back to pre-2007 levels. This may well be true, but a significant factor that has to be analyzed is: what jobs are being taken. I have no numbers only anecdotes [from UK/Europe, not US], and there is a pattern where low paying service jobs are replacing jobs in the manufacturing industry. This is a direct consequence of automation in the manufacturing industries, and these jobs are not very secure / automation-safe either. Also, this problem is way older than you or me, the first Luddites were smashing machines in the 1810s. The history of the 1800s show that automation (with lots of other factos in conjunction) can lead to societal upheaval and severe economic problems.
As for productivity growth, I concur that by the numbers in your first link show a stagnation in the past decade. That does not mean that this trend will continue in the future. When AI/ML Technologies become mainstream (this is already the case) and receive widespread adoption in several industries (this is not the case by now) productivity growth will increase once again.
We did indeed hit "Peak Oil" for conventional oil fields just as predicted. But we started to access new sources of oil to fill the gap. As long as we find new energy sources that are cheaper than oil we will never run out of oil because we can simply make more from other sources. Every form of good energy can be converted into every form of other energy (at some losses). Food can be turned into energy and energy can be turned into food. The big question is if we will develop cheap energy sources quickly enough.
If we don't we will have to reduce consumption considerably and there are several ways to achieve that.
I was referring to the quasi-apocalyptic belief that was popular after the Iraq War (oil was on the verge of running out, leading to resource wars and societal collapse). The Oil Drum was a good example of this[1].
>there is a pattern where low paying service jobs are replacing jobs in the manufacturing industry. This is a direct consequence of automation in the manufacturing industries
It's kinda more related to that "Made in China" sticker that appears on most manufactured things you buy these days.
No shortage of manufacturing jobs in China. Or Germany (which has also enacted economic policy to keep manufacturing at home), for that matter.
If prime age productivity were actually improving, then there would be more investments in it - instead firms are buying back shares, instead of re-investing.
So The marginal choice is known - perhaps the easy wins to be won from human productivity increases have been achieved?
> instead firms are buying back shares, instead of re-investing.
That's false. Firms have been spending more on expanding capex investment than stock buybacks. There's a capex boom going on with the S&P 500 right now.
Year over year capex growth was up 15.6% in the first quarter, and 18% in the second quarter.
By contrast, capex growth for the S&P 500 was flat to negative for most of the prior five years, and was negative for nine straight quarters prior to 3Q17.
We haven't seen that kind of capex boom since the initial bounce out of the great recession in late 2010-2011. And that was massively juiced by the Fed, whereas we're in a very different Fed environment now.
This shows that capex investments are only starting to ramp up, are unevenly distributed.
In general it supports the position that stock buy backs are the preferred route of money use given the current monetary scenario. Especially since stock buy backs are at record highs.
This helped me find another point of interest - that 10% of households hold 86% of stocks. Meaning that the gains to shareholders will only re-concentrate wealth.
To add to this point, over the last 12 months the US has created the most manufacturing jobs since 1995. Manufacturing jobs had essentially declined non-stop from 1998 to 2011 (there was no recovery during the real estate bubble years of '03-'08).
That was supposed to be impossible. It was seemingly universally claimed by economists, industry pundits, stray talking heads on TV and armchair experts that manufacturing jobs were in permanent decline.
Instead, the US is in the middle of a manufacturing boom. That looks set to continue with at least three factors in the US favor: cheap, abundant energy; the cost of manufacturing in China continues to climb and has made the US a lot more cost competitive in contrast; a slowdown in the DC regulation monster; corporate tax cuts are a big deal for domestic manufacturing - as profit margins in manufacturing are typically modest - and the US is now tax competitive with the OECD median.
To paraphrase Buffett, you get rich making things, and you get poor buying things. The US used to be a potent example of that; China is the latest stellar example of it. The US desperately needs to reduce consumption and boost production + savings, in regards to the ratio in the economy.
It also ignores human creativity. "Machines are better, so no more jobs", when the reality is "people like to pay a premium for hand-made things / services that cannot be provided by machines". Sure, there's no guarantee that we'll create new jobs, but it sure seems very likely to me that the market will find a way to employ people if the market is allowed to.
Let's look at actual measures of automation - productivity growth. It's been very low over the past decade[1] when compared to decades past. Prime age employment has been returning to pre-crisis highs[2] - despite people claiming that the old jobs were gone and new jobs wouldn't replace them (the machines were supposedly going to take them all).
[1] https://www.bls.gov/lpc/prodybar.htm [2] https://data.bls.gov/timeseries/LNS12300060