Summary: This post discusses some recent articles about automation and unemployment, and my reactions to them.
Following my post about automation and unemployment, I got several comments saying I was too optimistic, and (fewer) saying all this would never happen.
I am still trying to “get the future right” (I know this is really impossible). So here are some interesting related things I read lately (two pessimistic and one somewhat-optimistic). I promise to get back to verification in the next post 😉
Unnecessariat is a grim, sometimes heart-breaking description of the currently-unemployable. If we assume that automation will make many more people unemployable, then it makes sense to look at areas which have already passed from employment to mass unemployment.
And according to this writer, it is not a pretty sight. In the section “The view from here”, she says:
Its no secret that I live right smack in the middle of all this, in the rusted-out part of the American midwest. My county is on both maps: rural, broke, disconsolated. Before it was heroin it was oxycontin, and before it was oxycontin it was meth. Death, and overdose death in particular, are how things go here.
Was I too optimistic in my post? Perhaps the story she tells, full of overdose, suicides and a feeling of helplessness, is a more accurate description of a future-without-work? I hope not: I think we stand a chance to avoid it (with the security that comes with UBI, with changes of perspective and so on). But it is not going to be easy.
Tyler Cowen’s article about deindustrialization
In my original post, I specifically limited discussion to the results of automation in the developed world. But how will it influence the rest of the world?
Economic Development in an “Average is Over” World (pdf, long) is a partial answer to that, by economist Tyler Cowen. His main claim is that automation may cause developing nations to shorten or skip the industrialization phase. This is not good: Industrialization is really important for the creation of a healthy middle-class, which in turn contributes to democracy, income equality and a lively civil society.
Thus, in Asia, some countries (Japan, Taiwan and South Korea) went through a lengthy industrialization phase, and thus are now democratic and have reasonable income equality. China started later, and is already de-industrializing, and thus has very high inequality and no democracy. And some other places (the Philippines, Uzbekistan, Kazakhstan etc.) may find it very hard to industrialize in an age of automation, and will thus end up not very democratic and not very equal.
Here is some of his description of why industrialization is important for equality:
Commodity exports, however, typically lead to especially high levels of income inequality. If we look at the most equal economies in Asia – Japan, Korea, and Taiwan – they have little in the way of natural resources and they relied heavily on savings and human capital. Most of the major commodity exporters have quite inegalitarian outcomes, a logic which is perhaps most visible in Russia. It is relatively easy for political elites to gain control over the resources and capture most of the rents for themselves, as is often the case in oil-exporting countries. It’s not so easy to do the same for human-capital based manufacturing wealth.
But it ain’t all bad: Even without full industrialization:
Most of the developing world now has cell phones, and this may represent a new and different growth paradigm, namely one based on the immediate spread of consumption opportunities.
Or in other words, rather than Indonesia or Cambodia exporting manufactures to buy imported goods, an alternative development path is that some of those imports trickle down and enter poorer countries at especially low prices.
I find his description convincing – both the pessimistic and the optimistic parts. Note that his description of how everybody will enjoy the fruits of automation is somewhat similar to my description (in the above-mentioned post) of why huge income inequality within a country will cause only medium quality-of-life inequality:
Also, remember that almost everything at some fixed quality keeps getting cheaper: Even now, the guy who drives a car 10 times as expensive as yours does not really get 10x quality out of it, right? Depending on how you count, maybe he gets 50% more quality out of it (if that), because pretty-good cars became relatively cheap.
The Economist’ special report on artificial intelligence
Finally, this week’s Economist has a special report on AI (leader is here, full report behind a pay wall). They really talk about automation in general. It is a good summary (and the Economist is my favorite newspaper). Below I emphasize the points where I tend to disagree.
The economist starts by making the point that we have seen it all before, and people managed to find new jobs. It than makes the somewhat stronger point that while old jobs will be automated away, new needs will arise and with them new uses for people. Ignoring this, they imply, is simply lack of imagination: For instance, ATMs indeed reduced the number of bank tellers per branch, but also enabled opening more branches (because opening a branch became cheaper). Overall, bank employment did not fall.
Needs are indeed pretty flexible: In a separate issue of the Economist they made the point that improvements in lighting technology (LEDs and all that) may not reduce overall lighting energy consumption, because there is no known upper limit on how much light people will want at night.
They go on to say:
So who is right: the pessimists (many of them techie types), who say this time is different and machines really will take all the jobs, or the optimists (mostly economists and historians), who insist that in the end technology always creates more jobs than it destroys? The truth probably lies somewhere in between.
This special report has therefore focused on the practical effects of AI in the nearer term. These are likely to be a broadening and quickening of the spread of computers into the workplace and everyday life, requiring people to update their skills faster and more frequently than they do at the moment. Provided educational systems are upgraded and made more flexible, which is beginning to happen, that should be entirely feasible.
My intuition is that indeed, new and currently-unimagined needs will arise. And better (often AI-enabled) education will help many people answer those needs. But unfortunately a growing percentage of people will be left behind (because they don’t have the basic capabilities to learn this new stuff, because the changes are coming too fast, because machines can do these new things cheaper and more reliably, and so on). The Unnecesseriat article above paints a picture of how that is already happening.
About the dangers of Artificial General Intelligence they have this to say:
AGI is probably still a couple of decades away, perhaps more, so the debate about what it might or might not be able to do, and how society should respond to it, is still entirely theoretical.
I think the people who are worried about AGI are also saying “probably in 20 to 60 years” (see my summary). They just make the point that we’d better start working on this now, because the problem is so hard.
BTW, there has recently been some progress in AGI safety research, e.g. about Safely interruptible agents (aka “the kill switch”). The ability to interrupt advanced AI agents and “change their minds” is an important piece of the friendly-AI puzzle. The hope is that these pieces will eventually culminate in a solution to the AGI safety problem before we actually need it, but (unlike your typical Hollywood movie) there is absolutely no guarantee that a solution will arrive on time.
Finally, they say:
So far the debate has been dominated by the gloomy possibilities of massive job losses and rogue AIs. More positive scenarios, in which AI dramatically changes the world for the better, tend to attract less attention.
They give three examples of things which will change the world for the better: Autonomous Vehicles, advanced personal assistants, and AI helpers in scientific and medical research.
They are clearly right about the positive stuff, but I don’t know many people who are obsessed with just the “gloomy possibilities”. AI / automation will probably do all three: It will change the world for the better in many ways, it will bring a lot of unemployment, and it will bring (somewhat later) the AGI danger, which hopefully we’ll be able to avoid.
I’d like to thank Kerstin Eder, Ziv Binyamini, Sandeep Desai and Amiram Yahudai for commenting on previous versions of this post.