I have been frustrated recently with my inability to efficiently participate in discussions of automation which crop up online and in person. The purpose of the post is to refine a conversational presentation of what I believe to be the salient concerns; the chief goals are brevity and clarity, but obviously corrections of fact supersede this.

Epistemic status: plausible causal conjecture.

I think the current wave of automation will be different from previous ones, in ways which make it more disruptive. There are three reasons for this:

No Fourth Sector: The economy has three broad sectors: agriculture, manufacturing, and services. The first wave was in agriculture, and people could find adjacent work or switched to working in manufacturing. The second wave was in manufacturing, and people could find adjacent work or switched to work in services. The current wave is affecting services, but there is no fourth sector of the economy left for workers to switch to.

Skills Over Jobs: Agricultural automation was largely about tasks: a digging machine, a seeding machine, a pulling machine. Manufacturing automation took this to the next level, with robots performing defined sequences of tasks. But in both cases these were specific - any task or series of tasks which had not been specifically automated was still work to be had. The new wave of automation is entire skillsets, like apply this pattern or the ability to speak. This means when a job is lost to automation, all similar jobs are going away at the same time. There will be no adjacent work for people to switch to.

Speed: When automation was physical machines, they had to design them, and build them, and ship them, and customers had to rebuild their own factories to use manufacturing robots. Modern automation is largely software driven, so design and build are the same process, which is then practically free to copy and distribute. As soon as the method is ready, it can be picked up by businesses as fast as they can rent server space to run it. This gives local economies and institutions like government very little time to respond.

Automation is different this time because the problems we experienced last time will be more severe, and more widespread, and happen faster.

New to LessWrong?

New Comment
26 comments, sorted by Click to highlight new comments since: Today at 8:44 AM
No Fourth Sector

The way we happen to classify jobs shouldn't be used to predict the future. Some jobs will be automated sooner than others, even if they're all called "services".

This means when a job is lost to automation, all similar jobs are going away at the same time.

This seems also true for agriculture and manufacturing (when a cookie factory was automated, a cake factory was automated soon after).

Modern automation is largely software driven

This is false for self driving cars and many tasks that will require robotic hands.

The way we happen to classify jobs shouldn’t be used to predict the future.

The prediction isn’t based on the classification, it just conveniently captures the relevant information in a way we can expect most people to understand.

(when a cookie factory was automated, a cake factory was automated soon after)

This isn’t what I mean by adjacent work, but does a good job of illustrating that the sentence is unclear. How about this:

”This means when a job is lost to automation, all the jobs that person could do are going away at the same time.”

This is false for self driving cars and many tasks that will require robotic hands.

I disagree. Neither self driving cars nor robotic hands require any fundamentally new technology - there has been a large gap between the kinds of sensors and precision mechanisms we can build and the control software we require to use them. The only physical innovation required for self driving cars is battery capacity in the case of electrics. We have been able to remote control vehicles via radio since 1894 - what we needed was software to do the controlling for us. Consider that the traditional car manufacturers’ testing largely consisted of retrofitted regular cars.

That being said, your larger point stands - I think I should make some mention of the fact that extant automation techniques are still going to be working alongside the new ones.

The prediction isn’t based on the classification, it just conveniently captures the relevant information in a way we can expect most people to understand.

What relevant information? My point is that the sentence "The economy has three broad sectors: agriculture, manufacturing, and services" does not have any information relevant to answering "what jobs will remain after 20 years". If economists had identified four sectors, instead of three, this would not make automation any less threatening.

”This means when a job is lost to automation, all the jobs that person could do are going away at the same time.”

This is false. A truck driver may loose his job soon, then he might move to a job in walmart, and it could take decades before that is fully automated. Of course, if you believe that all jobs will be automated soon, then the statement is true in the trivial sense.

Neither self driving cars nor robotic hands require any fundamentally new technology

Your post didn't mention fundamentally new technology (I guess we could debate what technology is or isn't "fundamentally new"). I understood that you were talking about large costs and delays of adoption, caused by the need to manufacture and sell physical things. Many tasks that will be automated do require physical machines. A logistics company will need huge investments to make their trucks self driving, and it might take decades before most companies are willing to make them.

If economists had identified four sectors, instead of three, this would not make automation any less threatening.

That is exactly the relevant information I need to capture. The problem is that there are no large, underserved categories of value which unskilled labor can accomplish, where automation is not imminent. The three sectors is something everyone understands, and has continuity with the rest of the automation discussion (like popular books and articles on the subject).

A truck driver may loose his job soon, then he might move to a job in walmart, and it could take decades before that is fully automated

The distinction between having their way of life destroyed vs. having their way of life destroyed but probably not starving to death is overly fine for the population at large. Further this ignores the context: there’s a lot of truck drivers, and there are already more than enough people to work for Walmart (and all similar positions). Though I see I still fail to have reinforced the intuition about the automation of skills. Sentence Mk. III:

”This means when a job is lost to automation, all the jobs that person could do **with that skill** are going away at the same time.”

I understood that you were talking about large costs and delays of adoption, caused by the need to manufacture and sell physical things.

That is true, but the focus remains on what is different this time. I may be wrong, but I strongly suspect that no one views better robots as different in kind from the sort of automation we saw in the 1980s. This is just an incremental improvement over the kinds of things we could do before. What I want people to imagine are things like call centers (unskilled, but only recently automated because talking to people was hard for machines) and accounting or the law (which implies a high level of skill is not a guarantee). Do you suppose I could make this clearer by just citing the call center example? I am not confident that ‘renting server space‘ communicates the ease to people who have no familiarity with it.

The problem is that there are no large, underserved categories of value which unskilled labor can accomplish, where automation is not imminent.

This may well be true, but that has nothing to do with the three sectors. Automation doesn't actually happen one sector at a time. It's not like only service jobs are going to be automated soon. There are still people working in, e.g. agriculture, whose jobs will be automated too. Talking about sectors is completely pointless, it's a non sequitur. You'd be better off just using the sentence quoted above instead.

”This means when a job is lost to automation, all the jobs that person could do **with that skill** are going away at the same time.”

I need you to show me that this wasn't the case in the past. The farmer who was replaced by a tractor didn't find another job that requires plowing.

What I want people to imagine are things like call centers (unskilled, but only recently automated because talking to people was hard for machines) and accounting or the law (which implies a high level of skill is not a guarantee).

How many people are threatened by that kind of automation? Is it a significant number, compared to truck drivers, wallmart workers and all the other jobs that do require physical machines?

" That is true, but the focus remains on what is different this time. I may be wrong, but I strongly suspect that no one views better robots as different in kind from the sort of automation we saw in the 1980s. This is just an incremental improvement over the kinds of things we could do before. "

Here's what I think is different about it this time. Specifically, the 1980s methods involved basically : you identify the task to be automated. You hire an automation engineering firm to do the task. This would involve custom "fixtures", custom mechanical assemblies, and very carefully designed assembly lines. Then a team of programmers has to very carefully put together a sequence of commands to complete the task. If you binge-watch "how it's made" videos, you will see hundreds of examples of this. Cake factories where a dipping machine is custom made for the shape of the cakes going into the chocolate dip, where it can't make a different kinds of cake or change the chocolate ratio or make rice crispies instead or learn how to clear faults where the cake has clogged up the machine.

One big flaw with this is if you have a new kind of cake you want to make, you can't just send a description of your new recipe to the factory and have it make it. Nor can the equipment try different strategies for cake making and self-optimize for higher speed and less errors. Nor can the equipment be given a new set of robot hardware that has slightly different performance parameters and self-adjust to make the cakes using these different robot actuators. Nor can it avoid "hurting itself" by carefully planning each motion and making sure planned motions are not going to hit anything in the environment. Some human has to set all this up. Some human has to manually adjust timings, to unclog it whenever it faults, and so on.

Today's methods show all of the above is possible in the immediate future. Also, we can potentially build massive frameworks, where automating even small tasks is easy because the framework is stable, easy to use, easy to connect to compatible hardware. And you just go on the "app store" for the framework and rent the pieces you need to do a task. Go rent a classifier that can recognize most objects seen through a camera. A physics modeler that is both learning and already pre-trained for most ordrinary objects in a factory. Scoring software that can measure outcomes. Just grab all the pieces and put together your automation "app".

And obviously we can go to either waldos that can do many tasks and don't need custom mechanical fixtures, or we find a way to rapid-design fixtures and get them installed and working automatically.

And, hopefully, robots making cakes will be able to share knowledge back to the cloud, so that other robots elsewhere making transmissions get slightly better, or if an employee drops a cake on the factory floor, they know what it is.

So ultimately, while it'll take a long time to actually build all this, ultimately we can automate all retail stores, all warehouses, all farms, and all factories. Billions of jobs. I don't know how to automate service jobs like cutting hair with the current state of the art, but even if we can't, if half the population of most countries are out of work and trying to re-train to cut hair or something, it crashes the labor market for barbers, etc.

There wasn't a large "manufacturing" sector for agriculture workers to move into, it became a large sector as the workers moved into it. Perhaps some current small sector of the economy will become a large sector as workers move into it? At least in the U.S., there's little evidence to support your claims of it being faster and more widespread-- jobless rates are at historic lows. Unless you mean it hasn't yet begun.

All that said, though, it is certainly the case that if you have a robot that can do anything a person can do, you don't need to hire any more people, and there must be some kind of curve leading up to that as robots become more capable.

There is no hypothetical sector just waiting for workers to move to. Manufacturing had always existed and grown steadily so long as the food needs of the population were being met by fewer and fewer agricultural workers. The shift didn't sneak up on anyone.

Those "historic low" jobless rates are much lower pay and skill on average than 10 years ago. Over the last 50 years people in those low-skill jobs have (when adjusting for inflation) been steadily making less money. We've reached market saturation on workers needed to meet the demands of the population.

I'm one of those people automating things. The really scary part is how much more quickly those of my ilk can automate processes now than even a few years ago. Expect the drop in demand for human effort to get worse exponentially.

design and build are the same process

I don't understand what you mean by this. Software still requires both design and building, and they're distinct tasks. Design is defining the problem and deciding how to approach it: what classes to write, what libraries to use, what the deliverables will be, etc. Build is writing the code, fixing bugs, deploying it, etc.

You are right. This is much more an administrative concern than a physical one - ad-hoc software deployment is merely bad practice, in manufacturing at scale it is actually impossible - but then again I can't imagine automation software falling into the ad-hoc category. I am also skeptical of the value of bringing up how software is made at all, when the point is that it is really fast compared to physical products.

That is true, but remember that once the program is released there is close to no time before anyone can access and use it. Neither is there a limit to how many may use it simultaneously.

Compare this to the process which happens after a automated factory robot has been designed, or a tractor: The materials for the hardware has to be ordered, shaped, put together and tested. The unit has to be shipped to the customer who ordered it, in some cases it also has to be installed.

The lead time from first finished product unit to when the customer can use it is in orders of magnitude shorter when we're talking software. This lowers the cost for implementation significantly which should lead to a greater adoption faster, assuming similar amounts of productivity gain.

Technically, there is a fourth economic sector. The quaternary industry is what computer programming and research positions fall under. However, that doesn’t mean we can all realistically move to that sector, because it would require every person without a job to go through a masters or doctorate program, which is especially difficult to do when everyone who needs it just lost their jobs, and the quaternary industry is also being automated, so the point still stands that automation will likely lead to most jobs being wiped out.

That is very interesting - I had not heard of this notion before. Can you by chance recommend a source where they dig into the motivation for segregating this out into a new sector?

Naively I would be inclined to treat research and programming as the informational equivalent of farming and manufacturing, but it occurs to me the floor of abstraction could render it fundamentally different in some other way. It does not seem to change very much from a value-to-consumer perspective.

Moved to frontpage.

Actually I'd say the jobs best placed for avoiding quick automations are mostly jobs in the agricultural and building sectors - i.e. low skills* physical work. 
In a few years some Silicon valley start-up will discover that Indian or Vietnamese or Nigerian engineers/data scientists can do the same job at a fraction of the cost... and this will have no economic impact whatsoever since basically everyone working on a computer will be made redundant by GPT-N or something (without even taking into account AGI). 
But the guy who put his hands inside the engine of your self-driving car is here to stay for a long time (as is the one doing the pipes in your house or even the guys plucking out strawberries in a farm). 

 

*a lot of so-call low-skills are actually pretty challenging and require a wide variety of intellectual and physical skills. I'd recommand Shop class as soulcraft by Matthew Crawford on this subject.

Second round! Here is a more condensed version that might be appropriate for something like Reddit comments. The idea is that someone will ask the question "What is different this time?" and this will be the answer, with the expectation that it can be expanded in follow-up discussion as needed.

Nowhere Left to Run: When we experienced big automation waves before, there was a need in new and expanding industries for similar skills. For example, people went from farming to manufacturing, and from manufacturing to services. There are no industries like that now.

Everything, All the Time: Modern automation doesn't just automate a job, it automates an entire skill like apply this pattern or speak a language. This means that when a job goes away, all the work that person could do with the same skillset starts to go away at the same time. It also means that places we used to think were safe, like law and medicine, will be automated.

Faster Kill Faster: A lot of modern automation is software driven, which is very fast to produce and practically free to copy and distribute. This means less time for local economies to adapt, or governments to react, or workers to retrain.

Davos, which I suspected of being the modern Stonemasons (https://www.youtube.com/watch?v=XXtQMz1RGNw), actually looks a like a very positive organisation (promise you I have no connection). Hope nobody minds if I add a few quotes from...

http://www3.weforum.org/docs/WEFAnnual Report_2015-2016.pdf

Shaping the Fourth Industrial Revolution, although not a formal System Initiative, provides the contextual, intellectual framework for all of the Forum’s System Initiatives and related activities.

The world is experiencing unprecedented change, driven by the technological shifts of the Fourth Industrial Revolution. As ICT drove the third industrial revolution, digital as the output is the basis for new systems thinking that characterizes the fundamental shifts.

System Initiative on Shaping the Future of the Digital Economy and Society

Global events this year have emphasized the fragile nature of the post-war order, the systemic nature of decision-making and the weakness of our global governance framework.

Of particular concern is an acute moral crisis caused by a critical erosion of trust in leadership, in the ability and motives of experts, and in the systems that distribute our political, financial and human capital. We can confront this crisis only if those in positions of responsibility once again become role models for ethical behaviour. Here, the Forum will not be afraid to act with purpose and campaign for universal values. We will build our activities on a foundation of three basic human aspirations that the Global Agenda Council on Values has determined are widely shared across cultures, religions and philosophies:

– The dignity and equity of human beings – whatever their race, gender or background

– The importance of a common good that transcends individual interests

– The need for stewardship – in the sense of a concern for future generations.

World Economic Forum Annual Meeting 2016 (20-23 January, Davos-Klosters, Switzerland)

Over 2,500 leaders and experts from 140 countries were active in advancing publicprivate cooperation to address critical economic and geopolitical issues as well as solve global challenges, such as climate change and sustainable development. The 46th Annual Meeting was also an unparalleled platform for co-design, co-creation and collaboration to address its theme, Mastering the Fourth Industrial Revolution. In this regard, the programme addressed the potentially disruptive change that would emerge from future scientific and technological breakthroughs. There were over 400 sessions of which more than 90 were webcast or televised to inform the global public about the insights,
debates and outcomes from DavosKlosters.

Annual Meeting of the New Champions 2015 (9-11 September, Dalian, China)

Convened under the theme, Charting a New Course for Growth, over 1,700 participants from 86 countries participated in what has become the foremost global summit on science, technology and innovation. In his opening remarks, Chinese Premier Li Keqiang underscored the confidence in the “creativity and entrepreneurial passion of the public” as future drivers of growth and development. The programme focused on rapidly emerging technologies and the ways in which politics, economies and societies might be transformed by them. This in turn led to discussions about the emergence of a Fourth Industrial Revolution. Its transformational impact was clearly visible as a result of hands-on learning through live demonstrations of new technologies, such as artificial intelligence, euroscience and robotics.

Liability of software coders is rising.

One thing that needs to be considered is that the median age of the population is increasing, The UN predicts the median world population age will increase from 30 in 2020 to 40 in 2085. Therefore the pool of low skill manual workers, typically made up of younger generations, is shrinking. On the flip side is the expanding pool of older people who need support services. Automation is a good solution to addressing this change. The change will only cause disruption if automation occurs faster than the change in demographics can support it.

The other consequence we need to consider is that automation changes the balance between labour and capital as to their ability to produce income. Whereas previously the majority of income is earnt by people selling their labour, as automation progresses the income earning ability will significantly increase for people with the capital to invest in such systems. This is probably the most disruptive change and the need to look at solutions like basic income.

watch this video if you don't believe me

https://www.youtube.com/watch?v=kDqLIwgHf2U

Thank you for linking the video.

I think that Peter Stone is speaking to a different question than the one I am trying to address, but it is a reasonable confusion to make since he is speaking to the AI question around which this community is largely organized. I am talking about automation in the colloquial way it gets talked about in media - robots or software doing work humans used to do. The conversation as I witness it usually plays out like this: someone asserts automation is a problem, and then someone else responds (correctly) that big automation waves in the past resulted in bigger, more productive economies, and most people found some other way to contribute.

The question is, if it turned out alright before, why wouldn't it turn out alright this time? The thing that I feel is lacking, including from the usual economics arguments, is any sense of why it turned out alright the first time. This is important because that is where we can tell whether to worry about it this time.

One of the intuitions that we are working with here is that whether automation is a problem depends on the margins. Rather than a question of whether all people are affected, the question is only whether enough are affected. I think the answer is yes, and I think that because mostly the same processes are at work this time as last time, but in a way that won't let us roll with the punches like before.

It's also worth considering that this is for dropping into an already-in-progress conversation, as distinct from starting a conversation from first principles.

[deleted by moderator]

Complains about downvoting are going to be deleted and result in a warming. This is that warning.

The fourth sector may be Creativity/Innovation

Creativity/Innovation in an environment where ideas are copied in China and marketed before they even make it out of a Kickstarter campaign?

True but after full automation, no one nation will have an advantage. It may be that we move to a global system which serves humans as both consumers and innovators. Perhaps then we can address global warming, poverty, alien threats, the need to get off this planet and our current inability to co-exist in peace. The battle between forces is what has driven evolution so far - the interplay of ideas in forums such as this, and in society as a whole, will drive us forward at an ever increasing pace. Let the AIs show us the consequences of our innovations before we commit. Let's see what Davos comes up with...