Once in a while, I like to pop over to the Singularity freaks at Singularity HUB; they are the sort of crazy that I often enjoy — trying to figure out where we’re going, based on the trends of current technology advancement. They will get most things wrong, of course, but it’s also great to see so much optimism packed in a single site 🙂
Recently they have been discussing the end of millions of jobs thanks to advances in robotization and artificial intelligence, showing how things might not be as bad as they seem; a topic, btw, that Extropia DaSilva is being addressing since past May on her own blog, on a huge eight-part essay.
I’m personally a skeptic and think that there will be a trade-off at some point. Jobs will be lost by the dozens of millions until there simply will not be enough consumers to consume dozens of millions of products at ultra-cheap prices manufactured by robots. So at some point, something has to be done. On my comment on Peter Diamandis’ article on the costs of products and services plummeting eventually to zero, therefore creating a society where work is (barely) unnecessary, I unfortunately (as always) reached the internal size limits when commenting, so I reproduce it here:
For context, if you don’t wish to read Diamandis’ article: his argument, well-supported by statistics and graphics, show that the cost of practically everything is falling more and more, and eventually some things will become ‘free’ — like, say, taking photographs, which was expensive before 1998, but now every mobile phone, even the cheapest kind, has a built-in digital camera — therefore rendering the ‘need’ of dealing with masses of unemployed people moot: if food, housing, transportation, energy, etc. become effectively ‘free’, then even unemployed people will be able to survive without an income.
So we basically need a graph here: what will plummet first to zero, living costs or salaries (in the sense that unemployed people do not earn a salary, of course)?
To achieve the goals you describe — and I’m well aware that they are all achievable in the 20-year-timespan — means massively kicking people out of their jobs. Not only by the thousands, or the millions, but by the hundreds of millions. And now we’re not talking about jobs with zero qualifications, like factory workers on assembly lines. We are talking, as you so correctly pointed out, people with higher education that are trained to take care of the elderly with years of experience — who will become as obsolete as the switchboard operator.
So, in order to get ‘the rest of us’ cheaper living — and I agree this will happen — it means getting more and more and more people without jobs. And not only that, but they will be technically ‘unemployable’, i.e. they will have no chance to acquire different skills to suddenly become available in the job marketplace again. This did work during the Industrial Revolution, as zero-qualification farmers and helping hands joined the zero-qualifications workers at factories. So you had terrible social consequences because of the swift change, and the desertification of the countryside in favour of large factory cities, but this time it will be much worse, because clever AIs and robots will be able do to highly specialised jobs and kick not only nurses but even doctors and surgeons out of their jobs.
And don’t even think that so-called ‘creative’ jobs will survive. Perhaps a little longer. Today, it’s technically possible to create relatively simple AI algorithms (for the currently available technology, that is) that could be easily trained with existing architectural models in order to create buildings on demand. Buildings, after all, have to follow a lot of rules and constraints, and it’s easy enough to create a rule-based expert system. What is lacking is the so-called ‘artistic inspiration’ from an architect. That can be done by simply feeding thousands or millions of models and train a neural network — this would be added as an additional constraint on the engine. Call it ‘the artistic constraint’. So you can give up on all architects (and civil engineers, since all they do is adding up values, something that any computer-based engineering platform ought to do automatically, and with much higher precision as well). Any doubts that this will be the case? Such technologies are currently being employed by academics on their virtual archaeology projects. Look up ‘Rome Reborn‘ — it’s a few years old, but they have been able to reconstruct almost all of Imperial Rome, based on ancient ledgers which did mention, street by street, how many buildings there were, how many stories they had, how many doors and windows, what material was used (wood, bricks, stone, concrete…)… and from that you can figure out a lot of proportions (based on surviving ruins, and common ergonomic measurements) — enough for a non-sentient AI to generate a whole city that looks pretty much realistic and at the same time is firmly grounded in historic data. Imagine what you could do with much more detailed models and rules!
So… there will be a pyramid… where the bottom will be eroded by AIs and robots. This, of course, is not new; it has already started, in a sense, with the Industrial Revolution. But I think that it was really in the 1990s when we start seeing whole job classes to completely disappear, being replaced by technology. A decade ago, people without qualifications had either the choice of flip burgers or to join a call centre. These days, call centres are starting to use AIs and not humans — first at the pre-screening stage, but, as time progresses, they will replace more and more humans, until none will be left. And this will happen very quickly. Take a look at AliExpress, one of the world’s leading online shopping website. Based in China, where labour is cheap, right? Yes, but the Chinese are hardly fools — their front-line ‘help system’ is a chatbot. And let me tell you that the few times I needed to find some information (how to get pro-forma invoices, for instance), the chatbot had no trouble in helping me out. So, yes, even the Chinese, land of many opportunities and low salaries, are massively employing AIs and robots to drive their costs even further.
The pyramid continues to get eroded from the bottom up. One might claim that the jobs of the near future that will be higher paid are robot maintenance technician and AI programmer. This will only be true for a short while. Robots will quickly maintain themselves (or each other). While we don’t have super-geeky, Sci-Fi like robots doing that, a lot of factories heavily relying on non-cute-looking robots already have some form of automated maintenance to avoid having human technicians all over the place — humans will only get called when something goes seriously wrong, but, as time passes, and robots get better and better at detecting faults and of correcting them, there will be less and less humans doing even that. What about AI programmers? Well, when I left university with my degree — and that was in 1992! — some of my colleagues were already working on self-writing AIs. Really — no joke! At that time, all they wanted to do is having AIs doing the tedious job of fixing bugs. But, of course, to fix bugs, AIs needed to learn to programme. Obviously we are talking about very simple things, proof-of-concept prototypes.
But over two decades have elapsed. Today, we have things like the IBM Watson processing petabytes of raw data from websites to figure out trends and profiling consumers — and of course Google has their own technology as well, and we know how good Google is at figuring out what we want (and at displaying ads!). All this would even be impossible with human beings; we simply are not fast enough.
So… yes, the rate of replacing humans by AIs and robots is accelerating, but we’re not really conscious about it. The recent financial crisis (which I believe is not really over yet — we’re merely pretending it’s over to let the markets settle a bit) has shown one important thing: as companies had to cut their losses, many had no choice, in order to survive, to seriously invest in automation — call it ‘generation zero robotization’ if you wish. But the whole point was that somehow you had to keep as many humans out of the production process. Cut the labour costs, and you become a capital-intensive operation — which is so much easier to manage, since those pesky humans and their unions and their demands will simply be out of the picture. The outcome? We might have ‘survived’ the financial crisis or not, but one thing is for sure: millions and millions of people world-wide not only lost their jobs, but became unemployable. Their previous job does not exist any more; it’s not only the particular company A or B that has gone bankrupt; it’s a whole job description that doesn’t exist any longer. In my country, architects tended to employ designers to do most of the tedious tasks of drawing blueprints. During the crisis, so many architects lost their jobs, that the offices got them to work cheap to replace the designers — so you effectively started hiring jobless architects to do the same kind of work as designers did, for less money, but getting people with much higher qualifications; at the same time, of course, nobody draws blueprints with pen and ruler any more. People use CAD software and net cms software — possible the latest generation, which can draw object-oriented 3D models from scratch, full with rich metadata which can be used during all the stages of the project. And it’s precisely those metadata-rich models that can so easily be fed into an AI that will ‘creatively’ do models on its own and replace those architects…
What did designers do? Most tried for an early retirement, if they could. A few went back to school in their middle ages, trying to become architects instead, and eventually get their job back — but with higher qualifications and lower salaries. But that’s just moving up a little bit on the ever-eroding pyramid: architects will be replaced next. And while obviously top-notch architects — those at the very top of the pyramid — will survive for a while longer, almost all others will become permanently unemployable (I mean, why should you hire an architect if an AI can do the same job, fulfilling all regulation and, as a bonus, design the house exactly how you wish, instead of what the architect wants to design?).
This will happen everywhere. Think musicians. Those great artists at the top have training, talent, looks, and do a complex creative activity, right? Well, not really. A lot of so-called ‘artists’ just have looks, since what they sing is remastered automatically with a piece of clever software called ‘Auto-Tune‘. This software is two decades old — and pretty much all so-called commercial performers use it at some point in their career, because it will make them pitch-perfect. Well, that’s one side of things. The other is automatic composition. Yes, that exists as well; yes, it does all the arrangements for you, all you need is to supply a melody to start with. But soon you’ll simply be able to figure out, say, the top ten commercial successes of the week, feed their melodies into an AI to train it to figure out what the public currently likes, feed it through auto-composer, get a state-of-the-art synth with outstanding sampling to play it, get a mediocre singer who looks great and let her sing with Auto-Tune in real time, and the result is an instant ‘band’ which will play exactly what the audience wants — and you’re just paying one human to sing, wiggle her attributes around, and show off some skin. Guess what will come next? Gorillaz was the first totally virtual band (appearing in 2001) which was a great success, but they still relied on composers, players and singers to sit ‘behind’ the characters to do the work. But these days you could get away without all that. And if you just want to do some rap, I’m pretty sure that certain synthetic voices are more than good enough for that. So, yes, once a highly paid job, requiring talent and lot of training, can today be completely replaced by software. Just because totally virtual bands haven’t become mainstream — just a few had some success so far — it’s just a question of time. Artists age.
They gain weight (look at Ian Astbury today!). They break down on stress and anxiety, or, after a particularly hard divorce, get on drugs and have a premature death. And in the mean time they might have simply horrible tempers to deal with. Digital recreations, by contrast, are always perfect and eternal, and will not sign up for unions nor have ‘a bad day’ — they will work at your bidding forever and ever.
Let me tell you another little story, which was told me almost a decade ago by a top marketeer on one of the world’s largest cosmetic manufacturers. The company he worked for owns one brand of hair spray that is iconic since the 1960s. They had a wonderful model back then to promote the hair spray; it was an immediate success; everybody would relate the brand with the model; and sales were always rising, because people who became fans of that brand would be very faithful, even after years — or decades.
And here is where things started to be problematic. In the 1970s, of course, the model still looked great, perhaps with a slightly more mature look, but that was fine, because her faithful buyers would have aged as well. In the 1980s, she just needed a lot of makeup from professionals to look great in public — and, of course, a strict diet, carefully controlled, lots of exercise to keep fit and trim. In the 1990s, the model was now in her 50s, so it meant a lot of surgery combined with makeup. In the mean time, the faithful customers would also have aged a lot — but they were hooked on the image of a youthful model (and didn’t want to get a reminder of how old they now were), with which their children now identified; and their children’s children would also start to listen to grandma’s advice when it came to hair spray. So, yes, sales continue to rise and rise. The model was expected to be present in public, to do ads, and so forth, and it was clear that things were seriously becoming harder and harder.
In the 2000s, there was some help from the CGI industry — thanks to CGI, you could have whatever models you wish always with perfect hair under all conditions. In fact, that marketeer told me that by the 2000s, all cosmetic brands used CGI to ‘improve’ their products. Even the most natural-looking models would be CGI-improved, and, of course, their image on the cans of hair spray were heavily photoshopped.
But the problem with this particular brand line was that the model would regularly do some presentations of the product, getting in touch with the faithful customer, and on all places where she went, sales would always increase — as grandmas would bring their daughters and granddaughters to use the products, and get in touch with the model, and talk to her, ask for an autograph, and so forth. This was now problematic — while there was no problem in using CGI to ‘improve’ the looks, there was a limit on what you could do with a 60+-old lady who was supposed to look like she was in her 20s.
So the brand had no choice: they had to start changing the overall image.
What they did was hire a handful of new models, and start putting them on the hair spray cans — paying close attention to the batches. Shops and supermarkets would get a certain batch with just one of the models, and see what the sales were. They would rotate among all models, in several cities in countries, to see which of the ‘new’ image would sell more.
The results were shocking: no matter what model was on the can, no matter on what city or which country, all sales were lower than with the original model. This was a serious blow to the managers. Clearly people all over the world bonded so strongly with that cute model from the 1960s, that the brand was doomed to die with the model.
At this point, desperate for a solution, a very serious proposal was put on the table: why not simply replace the model with a digital image of herself? The digital image would never age. And the CGI people could use real humans as well as digital ones with the same ease — probably it would be even easier. They showed the overall costs, and they were not overwhelmingly higher than what the current CGI state-of-the-art costed anyway. So, yes, here it happened, the top marketeers of one of the top cosmetic companies were very, very seriously discussing the replacement of a human model by a digital image — forever — and not going back.
My friend told me that the discussion was taken very earnestly. At the end of the meeting, however, one thing became clear: the digital model could do everything but one thing: be on stage and address the fans. So, at the end of the day, they decided to send the seriously aged model back to the top surgeons of the world and ‘fix’ her as best as possible, and make sure that she would not be too close to her audience…
That’s all very well, but… in the mean time, another decade has passed… and the model is not getting younger. I can imagine that the company is really able to pay for all types of surgery and make that model look as good as possible, considering her advanced age. But… one day, no matter how sophisticated medical technology becomes, the simple truth is that model, one day, will die, like all of us. What then? My friend has left the company (and went to work for the competition) for many years now, so we don’t know how the story ends (or what happened to the almost-80-year-old model in the mean time!), but he was certain that they would be using digital models more and more in the future.
Anyway… long comment, but… the point is pretty much to say that a human-less world is not really ‘science fiction’. It’s starting to happen. Technology in all areas — even on those where people always thought that computers would never ‘take over’ — are kicking people out of their jobs. The trend will continue. And yes, as more and more people are out of the value chain of products, those products will become cheaper and cheaper. I might have some doubts that we will ever reach the mythical ‘zero’, but we might get close to it. Take the Internet as an example. We tend to say that it’s all for free, but it’s not — we’re paying a monthly fee. The point is that the monthly fee is tiny compared to the amount of data we have access to.
So maybe, in twenty years, people will be able to eat quite well for $20 a month, and buy clothes for $0.50 or even less; that doesn’t
sound so far-fetched at all, since I currently buy dresses on AliExpress for less than $5 — dresses that, not so long ago, for their quality, would probably cost ten times as much. I do agree that people in twenty years will have much, much lower costs of living (while there will always be huge tons of gadgets to buy at hideously high costs). But we will have much, much more endemic unemployment, and that will have social consequences.
Unless, of course, something is done about that. There are currently two serious alternatives. One, of course, is the universal basic income, in one of its many formats. As more and more people are out of the production lines (and that includes services), companies will become more and more lucrative, which means that they can be farmed for increased taxes, and this, in turn, can feed more people on an universal basic income. After all, you need consumers to get the ball rolling, and those consumers will need to have at least some spare change to be able to buy massive amounts of those ultra-cheap products. This, however, creates a population which will be totally dependent on the State for basic survival, and I don’t know how good that will be; but there might not be another choice, except letting those people starve.
The other serious alternative is being done by Japan. Curiously enough, they are going on the reverse route: instead of lowering the cost of living, in certain cases, it’s much higher than everywhere else in the world. And why? Because there is a close collaboration between government and the major companies. One thing they all want is to engage in total robotization. The government says: ‘fine, but you cannot fire older employees’. What this means is that the company does get more productive thanks to robotization, they have higher margins, and this, in turn, allows them to keep their employees who have, for all purposes, life-long guaranteed employment. And what do the employees do? In some cases, jobs that (currently) are not yet performed by automation, like… holding the door for you. Pressing the buttons on the elevator. These are possibly very degrading jobs (since we do have automatic doors and elevators…), but the whole point is that at least those people are not unemployed. They earn well. And that means they can spend a lot. As a consequence, products and services are in high demand, and the prices have risen. That’s ok, because salaries have risen as well, so it sort of works out. For example, the typical small local shop in Japan, which would be staffed by one or two persons in the West, have perhaps 7 or 8 employees, trying to make themselves as useful as possible. You feel overwhelmed in the way you are treated, and wonder how they can keep so many employees. The answer is simple: raise the prices. That allows you to hire more people with good salaries; and that means, in turn, that they can afford more expensive things.
You might argue that such a model also requires a lot of collaboration between government and the companies. Japanese have no qualms about that; they’re used to it. The government is always interfering. 🙂 So the choice is to go towards a more ‘socialist’ society, where the State essentially has to give people money for them to survive, because there is no way for them to get a job; or, the alternative that Japan uses, the government forces companies to keep far more people (and pay them well!) than they need, so they use technology to keep their costs as low as possible, in order to be able to afford all those people — the result is, of course, almost zero unemployment, even if it means that millions of people are doing ‘useless’ jobs, which are done by machines in the West, just in order to keep those people alive, since they aren’t qualified to do any other kind of job. And, of course, as automation in Japan grows and grows, more and more people have to do those ‘useless jobs’, just to keep the society with happy consumers.
Does this work? Well, I have no idea, I’m no specialist; some obviously think that Japan’s system is hopelessly antiquated and is not fit to deal with the changing conditions of the current economy. And the main reason it has so little unemployment is because Japan, with Germany and Portugal, is one of the countries in the world with the lowest birth rate. This means that less and less people are available to work, and even retirees are re-hired part-time at low wages to fill in jobs where no one is qualified and skilled to apply for. Germany, of course, is attracting immigrants to increase its workforce; Japan is very reluctant to do so. The result is an artificially low unemployment rate, which might not be directly related to the employment system they have, antiquated or not.
To be honest, I love to discuss post-capitalism. And I’m curious of what will be the model of the future: universal basic income, or companies being forced to hire humans to do useless jobs. I’m not aware of a third solution. Not even starvation is an option: you need to keep consumers alive, or else you won’t have a market to buy your robot-manufactured products!
Update: To keep the discussion fair, it’s appropriate to add some contradictory, so that one can evaluate different views on the subject and establish one’s opinion. Some believe that at least in the short term, AIs and robots will indeed generate more jobs, and the proof lies in analysing the decrease in the unemployment rates in spite of more automation. Here are five reasons why automation will not kill all jobs but actually create more. I’ve also submitted a comment there!