ThoughtStorms Wiki

Over on the AltMoneyTribe (somewhere between 2004 and 2006) I was discussing AutomatingMechanicalJobs with a guy called Vlad.

Original posting was here but probably long gone:

Here was his question:

I was curious as to what you think we should do about the rapid mechanization of general labour. I think it will probably be at least 30 years before most jobs are mechanized but even if we add a slow (say 3%) mechanization of jobs starting with low skilled and steadily moving up the skill ladder, we'll really begin to feel the crunch within a decade to a decade and a half. The US could have unemployment rates rivaling those of Europe and other socialist countries, and if it is that bad here how bad would it be in the socialist nations of Europe.

And here was my reply / essay.

I broadly agree that we're capable of automating most kinds of work but there are two assumptions behind the question that I disagree with.

1) Firstly, the assumption that there's one deterministic route for technological development which isn't itself influenced by political decisions and motivations. I don't accept this. What technology we get, which things get automated and how, is partly determined by what we can dream up, but also partly determined by which projects we choose to invest in. Or even which seem cool to us. These are social, political and economic decisions rather than mere technological ones.

We can, to a certain extent, choose productivity improvements which aim for deskilling and eliminating workers, or we can choose empowering them. For example, the original Ford production line succeeded through economies of scale and division of labour from a deskilled workforce. The Toyota production line succeeded through greater flexibility requiring a more trained and skilled workforce.

More ForcesOfProduction

2) The second thing I disagree with is that there's one scale of intelligence or skill, and it's simply low-skilled jobs that get automated away. That's clearly untrue. Doing sophisticated mathematical calculations is skilled work which demands greater IQ than massage-therapy, but it's far easier to automate maths than massage.

The corrolory of 2) is that how automatable a job is doesn't really depend much on how much skill (or IQ) it takes, but on other factors such as variability within the work and need for interaction with (and understanding of) humans.

And automation doesn't necessarily simply replace like with like. Intricate pieces that need a skilled craftsman to make by hand can be stamped out of a machine 30 times a second. Driving a taxi in New York requires less skill than the craftsman has, but is currently beyond our ability to automate (Although it's fairly easy to believe that taxis can and will be automated maybe within the next 20 years).

What I imagine happening is not that jobs disappear, but that they change to be more and more about supervising machines.

InTheFuture, we'll have a lot of jobs which look rather like "QualityControl", where people are watching the output of the machine for evidence of flaws. Or looking at a computer screen showing a summary of statistics about the pieces coming off the production line for evidence of patterns of abnormality. Or watching a "CorporateDashboard" which shows the sales performance of dozens of different products, and making adjustments to "correct the flaws" in the product range so that it sells better.

This doesn't have to be "skilled" work. With the right user-interface, steering a fast changing product range to keep tracking the shifting tastes of the market might be something like driving the car in Gran Turismo.

Except it won't be fun. It'll be as much a slog as driving a taxi for 30 years, if you're working for someone else and on a low salary.

I'm not just trying to score points or get in gratuitous digs here. I'm trying to emphasize something I believe is important. Automation will drive us all towards working at a more "abstract" level. Each of us will produce more and manage more complex automated processes - but that doesn't mean work will seem any different to most people. It won't be more interestingly complicated and intriguing. Because even these higher-level manipulations will be simplified and standardized.

If you shift from carving a piece by hand, to operating a cutting machine, you become more productive and produce more complexity. But producing a hundred pieces an hour rather than one doesn't require you to be one hundred times more skilled or have one hundred times the interest. It fact it's a lot more boring and less skilled. The same will happen as we move to managing more complex machines and processes. One of the prerequisites to automating a task is to rationalize it. Strip away the complexity and variation.

Many jobs (from your doctor to the customer support engineer to the guy who fixes your car) will essentially involve looking at the results of diagnostic equipment, judging what it means, consulting a database, and then recommending which machine or process will solve the problem. (What drug or operation will cure you, what part needs to be replaced.)

Everyday. Repetative. Standardized.

(See also TurkingEconomy)

People will do some creative stuff, although the definition of creative might not mean what we think it means today. A lot of creative jobs are going to be destroyed by another important force : amateurization. I think a lot of writing, teaching, music-making, painting, craft, cookery etc. are going to end up in the gift-economy rather than the paid economy. (Bad news for anyone who wants to get rich as an artist)

In fact, amateurization should be seen as just as destructive of jobs as automation is. People will be so hungry for the interesting part of work, that they'll do it for nothing. That's the big story on the internet : how much stuff is moving into the gift-economy and being given away because that's what's fun and people want to do. Programming? Certainly. Art and graphic design? Sure. Music? Oh yeah. Once we all have nano-fabrication plants in our garage, any sort of industrial design is going to be churned out by keen teenagers. You'll design for yourself, and put it on the net in the hope of getting acclaim.

Of course, that won't cover all the intellectual work : the painstaking laboratory work in biotech; testing software on twenty different configurations of hardware. But this, boring work, is also increasingly about configuring and supervising automated processes.

You see the point.

Fun, creative stuff will go amateur. People will get paid for the dull, repetitive supervision of the machines.

Now, you might think that we only need a few people running the automated processes and everyone else will be unemployed. But I don't think that's what's going to happen. It didn't happen in the past and it probably won't in the future.

A Marxist analysis can explain why : the capitalist is not trying to maximize the amount produced per person, but trying to maximize the profit from reselling labour.

Let's take selling lemonade as an analogy.

Suppose you want to get rich selling lemonade. You currently sell 50 bottles a day for a dollar each and make 10 cents profit on each one. Now, which would you rather do : work as hard as you can to squeeze 20 cents out of 50 bottles a day? Or work as hard as you can to double production and sales to 100 bottles a day while still making 10 cents per bottle?

Although the increase is the same, most people will go for the expansion, because they see that there are more opportunities opening up there. There are possible economies of scale. Possibly you can break 100 and go for 200. If any new technologies come along that can get another cent profit on each bottle, you've got that times 100 rather than 50. Etc.

Now remember. The capitalist is in the business of reselling hours of labour. And the same economics apply. The capitalist will want to maximize the profit made from each hour of labour sold, but also increase the number of hours of labour being resold.

Faced with the choice of trying to produce the same amount with automation substituting labour. Or more, with automation rationalizing labour, the tendency will be to go for more.

That's not to say capitalists are crazy. Or acting like internet bubble VCs who wanted to hire people as fast as possible. There are limits. But there is a tendency.

And in general, that tendency will push the capitalist rather than hiring fewer people, into trying to sell more stuff. And this means, workers are still going to be demanded in the process.

Just a side note. If the capitalist is going to tend towards maximizing throughput, this also means consuming other raw materials and energy faster too.

Everything I'm saying here depends on there not being an ecological crisis, a social crisis or an energy crunch. Such events could knock the process back anything from 5 years to 5000 years. I'm neither a doomsayer nor a blind optimist. All these things do happen (all the time) on small scales, and it's perfectly possible they'll happen on larger scales too. (Just as earthquakes come in all sizes, so do eco-disasters.)

We may get away with it. We may not. I don't think we can predict very well in advance. But I think it's worth putting-in measures to prevent the obvious problems we can forsee.

OK. Where are we?

People are going to be needed to supervise the machines.

Because ultimately, there are a couple of things which can't be automated away : human attention and responsibility.

People will be increasingly paid to give attention to things, and to take responsibility when they go wrong. Even very smart machines will need human supervision because a) the human acts in proxy for the customer. The human will be the customer's interface to the automated process, the customer's "representative", making sure the machine does what appeals to the customer.

And b) the human always has a stake, something to lose if the automated process goes wrong. Customers will want that lever over the process : someone to blame or punish.

Now, you may be thinking. "What's the problem? This is all good. Production is going up. People still have jobs. Why is there any need for politics? (Apart from avoiding the eco-disasters.)"

Here's where my point that the technology doesn't follow a deterministic line comes in. All the above can be true, and there's still room for choosing between different variations in social development.

Or let's put it this way. You may think that the "taking responsibility" for the machine sounds like the sort of thing an entrepreneur does. We can imagine one version of the automated future where everyone owns their own machine, and each owner-operator competes freely in the market. The ones who tend their machines wisely : who do just the right preventative maintenance, decide to produce the things which turn out to be popular sellers, make smart distribution deals etc will thrive and get more machines; whereas those who tend their machines badly and make dumb bets on what will sell etc, will be poor.


But we can imagine an alternative future where most people never get a chance to own their own machines. When new machines first enter the market they're expensive and the rich get them first. After which, as the rich start to benefit from the machines, they buy more, and hire people to work them - which means work to pay attention and take responsibility (get fired, lose performance related pay) if things go wrong.

A smart owner of dozens of machines can effectively hedge his own risk by employing the machines in complementary areas. Whereas the people working for him are stuck watching one machine, dependent on its success.

In other words, I think a highly automated future is going to look not very different from ours. Except with a lot more automation and more productivity.

Despite the automation and productivity, most people will work most of their lives, on fairly tedious configuration and monitoring of machines that they don't own, for as low a salary as the owners can get away with. Meanwhile, they'll have acquired more extravagant habits of consumption (of all the stuff produced by the more productive machines) and will get themselves deeper into personal debt in order to pay for it, which will in turn keep them at work.

The obvious, smart way out will be to drop-out; to decide they need to consume and work less. And some people (an increasing number) will succesfully do that. But it will be a small minority of the total workforce. It will also be the case that increasingly people will inherit debts from their parents or other family members. It may be relatively easy to avoid buying expensive consumer goods, but hard to deny your sister expensive healthcare services.