There’s been a lot of worry lately that robots will take all of our jobs. Should you be worried? Should you try to make friends with the robots so they treat you nicely?

omnix

This would be bad

Now, there’s a lot to say here, so here’s the tl;dr: no, this is only moderately worrisome. What you should worry about instead are:

Worries about automation go back to the beginning of the industrial revolution, two hundred years ago. But, with some major caveats, automation is good!  After 200 years,

  •  Life for the majority of people is far better. Before automation, 90% of the people lived by subsistence agriculture, one bad harvest or pestilence or war away from death. And those scourges came almost constantly.
  • Americans, as usual, focus on bad things in America, and don’t realize that these are boom times for most of the world. Global poverty is way down; it’s never been a better time to be Indian or Chinese.
  • Despite all the worries about machines taking our jobs— they haven’t. US unemployment is currently under 5%—  which is about as low as it’s gotten in my lifetime.
  • In general, pre-automation jobs sucked. There’s a tendency to romanticize lost jobs, but you really do not want to be a cotton picker, or a miner, or a laundrywoman, or a data entry typist.

The thing is, at any point in the last 200 years, an alarmist could concoct a tale of machine devastation. With modern farming techniques we don’t need 90% of the population to work on farms. Omigod that means 90% of the population will lose their jobs!  Only, this didn’t happen. Only 1.4% of the US population works on the farm today; the rest of the 90% found other jobs.

Now, the major caveat: this process sometimes goes smoothly, but sometimes is hella disruptive. It’s not pleasant when a middle-aged person has to change careers, whether it’s an 1800s agricultural worker, or a 1980s steelworker. Whole regions can be devastated and not know how to pick themselves up.

Jane Jacobs had a lot to say about what happens when the process goes well, and when it doesn’t. She calls the successful places city regions; as the name implies, these are always near big cities. In brief, this is the belt round a city where automation produces new opportunities as fast as it erodes old jobs. In a city region, there is new work to do, and it doesn’t take a lot of intervention for people to find it. (The books on India I recently read are also good introductions to this process. Poor people are amazingly entrepreneurial when they get the chance.)

You can’t count on everyone to live in a city region, but you can manage the disruption in other ways. This is where you need a strong economic safety net. You want people to be able to change jobs.  It’s not a huge exaggeration to say that the New Deal succeeded because it cushioned the disruption of industrialization. Stimulus spending spurred production and job creation; Social Security allowed people to move to where the jobs were without abandoning their old folks; unemployment insurance kept people going between jobs; the GI Bill trained people for new occupations. Europe went farther, with universal healthcare and free university education.

(Do you want a universal basic income?  Go for it, so long as you’re not actually looking to reduce government benefits. But it’s a good idea on its own; there’s no need to drag the robots into it.)

OK, but aren’t the robots different this time?  They can drive cars now! They can take your order at McDonalds! Surely all the jobs will disappear!

The first thing I’d point out is, extrapolation is a crappy guide to the future. In 1890 you could predict that the cities of the future would be buried ten feet deep in horse manure. This didn’t happen.

Second, universal AI is a huge assumption. If you look at sf and pop-sci articles, humanoid robots are ten years away, and have been for a hundred years. The first robot story appear, Karl Čapek’s RUR, appeared in 1920. Basically, intelligence is a pretty hard problem, and researchers always underestimate it. It’s easy to feel (as I did when I was an undergraduate) that a pretty good AI would be just a few semesters of work. Well, it isn’t, or it’d be done by now!

Also, I spent years as a programmer, so I know just how stupid computers are. They are great tools, mind you! But I don’t think we should scare ourselves about their abilities, at least not yet.

The better question is, what sort of jobs can computers or robots do? The general answer: jobs that are

  • repetitive and predictable
  • expensive

Automation is not, er, automatic. It takes analysis, programming, and testing, and someone has to pay for all that. That’s why a repetitive assembly-line task, done by a high-pay union worker, is the first candidate for automation.  It’s barely worth it to replace a waiter (especially since they can be hired for far less than minimum wage).

(Driving is a weird case. I think AI driving is far less advanced than it seems. As in much of programming, you can cover 90% of the work of the program and still only be 10% done.  The unexpected or difficult cases take most of the effort.)

Let’s put it the other way. What jobs are probably safe from automation in this century? Some of these, I’d wager:

  • teacher
  • physician
  • nurse
  • CEO
  • programmer
  • athlete
  • writer
  • comics artist
  • prostitute
  • craft brewer
  • video game designer
  • marketing & sales
  • legislator
  • soldier
  • actor
  • day care worker
  • hair stylist
  • product designer
  • scientist
  • thug
  • organic produce farmer
  • architect
  • call center operator
  • plumber
  • robot designer
  • robot mechanic
  • robot debugger
  • cook
  • valet
  • monk/nun
  • preacher
  • personal trainer
  • psychologist
  • web designer
  • lawyer
  • burglar
  • drug dealer
  • cop
  • spammer
  • SEO farm writer
  • AI researcher
  • anti-AI pundit

Many of these jobs, though not all, involve what humans are best at: dealing with humans. I don’t think anyone cares that their cotton be hand-picked. I think it’ll be a long time before there’s a robot you would entrust your one-year-old to all day.

I have a friend who’s an architect. I’d say his work is at least half talking to clients, and managing building projects— i.e., managing other people (contractors and inspectors). There’s that human thing again. For making the actual plans, he already uses a computer. He can already produce a plan almost as fast as he can come up with an idea.

So the better question is not “Could a robot entirely do this job?” but “What could a computer-assisted person do in this job?” Lawyers, for instance, are often still stuck in the world of paper. Automation would allow them to take on more cases. (For good or for evil.)

I’ve purposely included some “bad jobs” on the list, because the point isn’t that “things will be fine.” But I’ll get back to the grim meathook future below.

I haven’t tried to anticipate what the new jobs of 2100 might be, but we can expect that there will be plenty of entirely new things. Over 200 years, we’ve moved from an agriculture economy, to a manufacturing economy, to a service economy.  I’ve suggested before that what’s next is a frivolity economy.

Another point that I think worriers-about-robots miss: Robots and programs cost money. As one datapoint: Bitcoin mining presently consumes as much energy as the entire nation of Tunisia.

Plus, if you’re really pessimistic about the uses of humans— then the cost of hiring a human will plummet. Humans can be raised quite cheaply, without the use of high-cost metals and rare earths, and they’re really pretty versatile.

I’ve written before about why humanoid robots are a dumb idea. I realize that many people really want them, but I’d answer that they only think they want them. You do not actually want a sentient android to be your sex worker, household cleaner, or driver, precisely because a sentient android can do what it wants, not what you want. Maybe you want a robot you can talk to— but speech is a terrible medium for giving technical instructions.

We’re way too influenced here by science fiction. We grew up thinking of the Jetsons’ robot maid, or C3PO. In fact, a bulky robot maid holding 19th century tools in her 21st century manipulators is awfully poor design. Consider all the household automation we already have: dishwashers, microwaves, vacuums, washing machines. Not a single one of them is humanoid, not a single one does its tasks as a human would. Honestly, automation of the house is almost done, compared to the year 1900. But if you want more, a better model would be the room-cleaning bots seen in The Fifth Element.

Here’s another way to think about the whole situation.  Again, 90% of the population used to be engaged in subsistence agriculture. That basically means that the entire population can do what the 10% did before. Or to put it another way, there are 325 million Americans. One way to explain our economy to someone from 1800 is that we’re as rich as a country of 3.25 billion people would have been in their time.

If we continue to automate predictable high-repetition tasks, maybe another 90% of current jobs disappear.  But the population will live like today’s 10% do. Their standard of living will be far higher, and their jobs on the whole more interesting than today’s. (Of course we’re writing sf at this point, so you’d might as well look at my attempt at an sf future.)

That doesn’t mean we won’t have a grim meathook future. Piketty has warned that our future might look like… the 19th century, when most income and wealth went to a tiny class— and not even a class of innovators and entrepreneurs, but a useless rentier elite. And of course right now as I write, a clown car of reactionaries is trying to take away tens of millions of people’s health care, while the clown-in-chief is demonizing trans men and women in uniform.

But that’s the thing: grim meathook future is a political choice. Automation is just a form of productivity increase— and productivity gains do not have to go entirely to the rich. They used to help out everyone.  Around 1980, American voters decided to stop helping out everyone, and help out only the top 10%.

If that continues, the future will be grim, robots or no.  But it’s not the nature of automation that is the threat. It’s whether we manage it under plutocracy or not.

 

 

Advertisements