Which is a shame because coming from The Economist, the original piece is expectedly bullshitty. The main claim is that while in the recent past (notably until 1975 but a discussion of the neo-liberal revolution would be outside the scope of The Economist’s blinders), wages increased basically continuously, this has changed since the onset of automation. To make its point, the article includes the the following claim:
Even in places like Britain and Germany, where employment is touching new highs, wages have been flat for a decade.
Notwithstanding that unofficial unemployment in the UK is currently higher than in the pre-GFC era, when it was higher than before the neo-liberal revolution. Notwithstanding that the German “Jobwunder” includes a low wage sector in excess of 20%, and even with that the official unemployment rate is higher than in the sixties and seventies.
This is something that Naked Capitalism should have called out, given that it had a post on the working conditions at Amazon Germany not too long ago.
The low-wage sector statistic explains the stagnating wages better than
Recent research suggests that this is because substituting capital for labour through automation is increasingly attractive[.]
especially given that instead of unionized Western labour being replaced by automation, much of it has been replaced by non-unionized Eastern labour. But yeah, this is The Economist…
In fact, this is something that the article admits itself when it writes:
As David Autor, an economist at the Massachusetts Institute of Technology (MIT), points out in a 2013 paper, the mere fact that a job can be automated does not mean that it will be; relative costs also matter. When Nissan produces cars in Japan, he notes, it relies heavily on robots. At plants in India, by contrast, the firm relies more heavily on cheap local labour.
So, actually it’s not “the robots” that are coming for anybody’s jobs, white collar or not, but the capitalists that try to increase profits. This is something that commenter Tim Gawne nails but that somehow escapes both “Unconventional Economist” and Naked Capitalism.
You know what’s not in the Economist’s tale? Organized labor! Instead, technological advances and increasing automation somehow magically bring greater prosperity, which is why economists
take the relationship between innovation and higher living standards for granted[.]
But that it took
until the “golden age” after the second world war that workers in the rich world secured real prosperity, and a large, property-owning middle class came to dominate politics[,]
due to strong unions who fought tooth and nail to make sure that some of the increased profits ended up in wage earners’ pockets, goes unmentioned.
Instead, labor’s victories are claimed to be the main reason why automation (or unprotected labor) undermines wages:
Today, because of measures introduced in response to, and to some extent on the proceeds of, industrialisation, people in the developed world are provided with unemployment benefits, disability allowances and other forms of welfare. They are also much more likely than a bygone peasant to have savings. This means that the “reservation wage”—the wage below which a worker will not accept a job—is now high in historical terms. If governments refuse to allow jobless workers to fall too far below the average standard of living, then this reservation wage will rise steadily, and ever more workers may find work unattractive. And the higher it rises, the greater the incentive to invest in capital that replaces labour.
See? It’s not lack of power for wage earners that causes capitalists to replace them with cheaper alternatives, it’s too much power! If we only allowed people to fall into poverty as soon as they don’t have a job, we could have the same unregulated low wage paradise as in China and Bangladesh. The below-poverty line Hartz IV “income support” in Germany and the sanctions on unemployed refusing shitty jobs are supposed to remedy this and, boy, have they delivered, as the Economist sees it.
And even if nominal wages went up, tough luck, because
However, prices of things not made of widgets, most notably college education and health care, have shot up.
Someone without blinders might point out that this is very much a US phenomenon and that in Germany for instance, it has been the below-productivity growth of wages more than the increase of education costs that has driven real wages down. But then the Economist couldn’t blame his favorite scape goat:
It is the increase in the prices of stuff that isn’t mechanised (whose supply is often under the control of the state[…]
See, not the underfunding of universities or their being run for profit is to blame, or the for-profit health care sector (both insurers and providers). No, it’s the evil government, notwithstanding that it delivers for far less cost in other countries.
Again, this is something that I would have expected NC to call out, and again it falls to a commenter, Jane, to do this job.
So the Economists narrative of how technological change impacts production and jobs is deeply flawed. In addition, I am not at all convinced by its claims and interpretations regarding what can and cannot be automated. I do academic Data Mining and Machine Learning research, which means that unless the private sector is vastly more advanced, I have a pretty good idea what’s currently possible and in light of this, the Economist’s examples seem to be a mix of white-collar but cognitively low-skill, and unrealistically optimistic (or pessimistic).
The Economist refers to a paper that supposedly claims that 47% of jobs are threatened by automation, which seems an oddly concrete number given that the repercussions of new technologies are difficult to foresee. In particular:
That includes accountancy, legal work, technical writing and a lot of other white-collar occupations.
At first glance, this is one occupation involved with numbers and two with text. Parsing and extracting numbers and text are two things that computers have basically been designed for but at least for the two textual jobs there would be aspects of understanding involved that current algorithms cannot perform yet. And the Economist goes somewhat into detail:
Text-mining programs will displace professional jobs in legal services.
Text mining algorithms can’t understand the texts they mine. At best, they can identify the parts of a text that are probably the most relevant, potentially in response to a query. The Economist gives Watson as an example but while Watson is a tremendous achievement, it is also a huge purpose built system fitted to a particular scenario. It has vast databases and non-deterministic inference rules that help it identify the relevant connections. So a similar system could maybe replace someone who looks up relevant laws or precedents w.r.t. a case, or someone who summarizes newly arrived legal texts. As a counter example to the Economists scare mongering, I’d offer the non-existence of the so-called semantic web. Google, with their interest in such a solution, with their R&D budget, and their proven success in past innovations, has so far not managed to roll out a search engine that goes beyond key word and phrase matching.
Biopsies will be analysed more efficiently by image-processing software than lab technicians.
Depends. In some areas, image recognition is already rather effective, so this might be true. Again, this is an example of white-collar, cognitive low-skill. This is about seeing patterns in an image (and btw, humans tend to see way too many patterns), this doesn’t require much conscious thought.
Accountants may follow travel agents and tellers into the unemployment line as tax software improves.
But that’s the thing: as “improves” gives away, such software has existed for years because tax regulations are essentially formal rules and what the program has to do is optimize the assignment of numerical values to different categories so that in the end the least amount of taxes are being paid. An accountant that advises very wealthy individuals or corporations essentially just extends this by considering optimization routes, such as creating subsidiaries, that the simple software doesn’t consider yet.
Machines are already turning basic sports results and financial data into good-enough news stories.
Once again, numbers, potentially augmented by some descriptions of how they came to pass – think play-by-play descriptions – are turned into full text. This is reporting. What this is not is journalism. An algorithm that can do this job would be incapable of doing investigative reporting. In the end, it only puts the information into a slightly different format – one that consumers might prefer.
The thrust of the Economist when it writes
but will there be many journalists left to care?
is therefore just more bullshit. Reporters might be replaceable but if journalists disappear, this will not be due to Machine Learning developments. Instead, it can fully be laid at the feet of business decisions that replaced true journalists with pure reporters and the occasional propaganda mouthpiece.
The Economist also shows the following table:
Apparently this is a reproduction from the paper they refer to. At first glance: why exactly should dentists not be replaceable? I’d assume that image recognition should do pretty well at identifying caries, and a robot probably drills and fills more precisely than a human can.
At second glance: economists only at 43%? The way I see it, there are two tasks that an economist can perform: prediction and prescription. Mainstream macroeconimists have a brutal error rate over the last five years, and as a predictor, an economist is essentially a serial learner/predictor: based on the current state of the world (and maybe some prior states) it predicts the next one.
There’s a very simple serial predictor: predict that the trends of the current state of the world will continue. W.r.t. the Eurozone, this primitive predictor would have outperformed the IMF, the OECD, and the ECB economists. So replacing mainstream economists in this regard should not be too difficult.
Furthermore, since much of current macroeconomics is based on mathematically rigorous models, making prescriptions what needs to be done comes down to numerical optimization, e.g. computation. This would of course be the wrong advice but not worse than what humans produce at the moment.
Proper prescription, however, involving an understanding of the economy, of distribution, of the society, of psychology, will be hard to model and hard to automate.
Not only is the argument regarding eventual wealth increases by automation bullshit, but the argument regarding why jobs are replaced and which ones is bullshit as well.
And “Unconventional Economist” is not helping. He’s agonizing:
Take, for example, the autonomous driverless cars under development by Google. They are a prime example of a how a human worker, such as long-haul truck and taxi drivers, could soon be replaced by machines[.]
Here’s the thing: that’s exactly the kind of jobs that should be replaced by automation, as commenter washunate also points out. Long-haul truck driving is a shitty job from everything I’ve ever read and heard, brutal hours, spirit-deadening work, low pay. In the same way it’s good if a lab technician has better things to do than stare at an image trying to identify abnormal tissue, if someone in a legal firm is not called upon to browse through law texts to identify the relevant ones. It’s waste to have humans do this if there is more relevant work they can be doing, as commenter rusti points out. And if the economist asks
Will there be airline pilots? Or traffic cops? Or soldiers?
the answer is clearly: if automation can make things safer and liberate humans from mindless, dangerous tasks, more power to it.
The truly relevant question, and the one the Economist doesn’t want to discuss because then it would have to show its true colors is what to do in response to this liberation, as commenters diptherio points out.
We can either swell the ranks of the unemployment and increase profits for those that already own too much. Or we can decide that human brains can be put to better use than image recognition and text search, and maybe along the way introduce those short working days Keynes envisioned almost a century ago.
And btw: a funny thing happened with all this automation. My favorite illustration is train stations in Germany in which counter personal has been replaced by vending machines…and now there’s always a few people hanging around there explaining customers how to use the machines. Or metro stations that have vending machines but still someone who advises you how to best navigate the network to get where you want to. Or the fact that for all self-selecting of books on Amazon or flights and accomodation for vacations and other trips, it comes down to humans in the end that help you find this author you half-recall or suggest something of a similar tone, or direct your eye to the sights less seen that make your trip special.
This is a long post just because of how much was wrong with the Economist’s article and “Unvonventional Economists” reception.
NC has the tagline “Fearless commentary on finance, economics, politics, and power” and I thought that this commentary was supposed to come from the postings. But once again it seems that it’s actually coming from the commenters – seems I’ve misunderstood the site’s purpose.