The robots are coming
It’s the strangest situation.
We are right at the cusp of a world, where we don’t have to work. Endless rows of shiny robots and blue led-lit servers might free us from drudgery for good. Yet a lot of people can also see the blood-stained metal spikes of the Mad Max buggies coming our way. Where does this schism in our pictures of the future come from?
We will take a look at this question in our sixth installment on justice!
A part of the answer is captured in this quote of one of the 20th century great physicists:
It’s hard to make predictions, especially about the future. Niels Bohr
What makes us disagree about what the coming industrial revolution will bring is that we are lazily extrapolating our current systems and habits in a straight line and change only few or even only a single factor. Don’t get me wrong. That’s probably all we can do. We don’t have any idea about the interactions of too many moving parts upon each other in complex, noisy systems anyway. It’s just too complex for us.
Depending on what we assume to be true about the current state of the system, our predictions will widely diverge. “Elites exploit workers!” and here comes the black smoke bellowing out of the Cybertrucks. Faulty batteries, one has to assume.
“The market is an evolutionary system to test ideas that provides unbelievable value to the people and expands our capabilities as a species!” and the buzzing of shiny robots can faintly be heard through the fog of the future more clearly.
You might say: “Dude, there aren’t any smart robots yet! Machines are still stupid! They fail often! Why even think about that?!” Let me start with agreeing with that. It’s true, there is no AI yet and that automation fails often. Here comes the “but”:
The problems to come
I am deeply convinced that it doesn’t need “artificial intelligence” to have a great impact on jobs. Stupid artificial workers will already have a great impact, because, let’s face it, a lot of our work does not really require a “natural” intelligence in the first place. And did you notice that humans fail all the time as well?
Robotic Process Automation, which is software that can do certain tasks repeatedly via the graphical user interface (GUI) instead of via an application programming interface (API), could replace millions of jobs by making the remaining ones more efficient. You have to perform the task once – open a mail, “CTRL+F”, search for “invoice number”, “CTRL+C”, go to Excel, “CTRL+P”, “Tab” – and the RPA can do it after that. If you have ever worked even in proximity to a cubicle, you just know how many people are doing stuff like that for a majority of their time.
Physical robots in tandem with humans are looking to be the best option to boost efficiency of workers and output of factories. The worker uses online teach-in techniques to show a robot how to do a task and the robot will do it without fail – and without any explicit programming. Small changes in the user interfaces have unlocked huge efficiency gains and speed ups, because programming can become a bottleneck in a lot of automated production process. A few large firms, like Volkswagen, already struck agreements with the unions that allows them to replace retiring employees with more automation. Unions and management agree that they have to become more efficient. And once the vocal, well connected, experienced employees leave for good, they can take the necessary steps in a piecemeal fashion. The demographic change is changing both sides of the feedback loop: removing friction to more automation and making it simultaneously more necessary.
And then there is something, I believe a lot of people don’t see coming yet: the advent of B2B services. If Amazon is capable of hiring, managing and paying employees all over the planet, why would they not be able to offer HR services globally? They know how to deal with local laws, regulations and unions. They know how to write valid contracts in local languages. They know how to comply with data privacy laws.
You know how AWS moved data centers, which were once the treasure of any company, to an amorphous “space” called “the cloud”? Why? Because AWS promised to care for the maintenance at a reasonable cost. And they did. I think the same principle of superior efficiency, less hassle for a company and easily predictable prices, will change what we call service jobs today into automated services that you buy on a platform.
And who do you think will be more aggressive in deploying and A/B testing the latest technology to their processes and systems to drive down cost further? Your 200 people company or Amazon, when it’s managing 3 000 000 000 people in their (hypothetical) HR systems?
Amazon already got all the computing and storage power they need for that. As does Alibaba, Microsoft and Google. And anybody who lacks them, can buy them there.
What will all those “SAP Specialists” and “People Experience Managers” do then?
Learn to code? Well, automation doesn’t stop there. There are a lot of tools that pair an easy to use language or even graphical tools with efficient, automated code generation in high-performance languages that require special training and experience to use.
All of that is without what we call AI.
I would (currently) classify “AI” as a collection of various statistical schemes (algorithms) that implement systems that get better with more usage and without additional human effort. You have your data (the more diverse the better it is) and your algorithm and build a model with it that you score with respect to some metric.
Roughly speaking, the more data you have and the larger the model gets, the better it tends to perform. It’s a huge simplification, of course, but I think it’s close enough to true to be instructive. Put in a different way: the largest user will get better faster than any competition. You see, why I think that kind of business makes a B2B platform especially interesting? Every time there are network and scale effects, it becomes a “the winner takes most” or even “the winner takes all” economy. B2B has both characteristics.
The growth in the capabilities of AI is driven by improvements in hardware and software. For a lot of problems we are still not certain, whether our current software paradigms, like deep learning, are enough to solve them all. Gary Marcus argued against that proposition in his recent book “Rebooting AI”. Deep Learning models are correlation machines. They lack “knowledge” of causal, spatial and temporal constraints and can therefore produce strange behavior. He thinks we have to do a whole lot more on the theoretical frontier, on the software side, if we want to fix the current flaws and get to “real” AI.
But if we increase the amount of data and the hardware resources keep on improving exponentially, an increasing number of problems becomes feasible to solve. Like playing certain games. Or driving.
Imperfect, yet better than humans or at least sufficiently good and cheap technology will be enough to be noticeable; a self-driving truck is not necessarily the first picture that comes to your mind, when you think of “AI”, but it could displace millions of jobs in the U.S. alone. There are more than 3 million Americans that are classified as “professional drivers”. Will the AI truck get into accidents? Almost certainly! Does it matter? Well, if it is statistically 20x safer than a human, I want that on the streets!
The models are sometimes made in a certain way that makes it really hard to tease out why they failed. Specifying all scenarios, where it could fail, is almost impossible. To bridge that gap until our models are good enough, there are also new hybrid approaches, where AI does most of the boring, easy to automate routine stuff and the human workers take over, when it becomes necessary. Remotely. From a comfy chair. You have all seen the military drone operators in their control centers waging war on terrorism somewhere in Afghanistan? That’s available for trucks, planes and ships.
To summarize: There is something to the Marxist idea of “dialectics”:
Enough quantitative changes can lead to qualitative changes.
There will be pressure put on jobs, that I don’t think is comparable to prior shifts in technology. It looks like this has another quality. The earlier industrial revolution was about replacing brawn with steam, this one is about replacing brain with electricity.
You could teach almost anybody to do small, repetitive, broken-down tasks in the factories of the 19th century. Do you think we can retrain enough people to become nuclear physicists, computational engineers, data scientist and designers to make up for that? We might be, but I am skeptical.
The effects
How bad would it be? “By 2015, automation had already destroyed four million manufacturing jobs, and the smartest people in the world now predict that a third of all working Americans will lose their job to automation in the next 12 years.” [YANG]
Gulp!
So, let’s say some – or all – of these trends persist. What will that do to our societies, if there are millions of people without the skills to compete against automated tasks? Millions of people who don’t have incomes, don’t pay taxes, don’t save for retirement, don’t buy anything. What do we do with these people? What will you or I do, when automation comes for our jobs?
There a different possibilities: the most dystopian one is where a rich elite, augmented by AI and implants, armed with drones and robots, decides that they don’t need the poor anymore. Season with eco-fascist (“too many people pollute the earth”), social Darwinist (“the cyborgs are the fittest”) or racist ideas according to taste and you got yourself a real hell on earth, where the poor are hunted for sports and exploited for fun. The Mad Max buggies of the introduction.
On the other end of the spectrum is pure bliss.
Everything gets created by communally owned and operated, eco-friendly and inclusive machines and all are supplied according to their needs; wars and wages are only a distant memory. The capitalist exploiters have presumably been brought to justice. Or something.
I think both scenarios are unlikely for the relevant timeframe, but I think we should ensure that we are as far away from the former scenario as possible. Our current systems rely on taxes levied on wages. When there are no wages, our systems will not work, how they used to work. So what alternatives are there? What do we have up our sleeves?
Universal Basic Income
Let’s start with what most people will have heard about by now. There was a whole presidential campaign run on the concept of a UBI by Andrew Yang.
A UBI is a type of social security that is guaranteed to all citizens in a country. It’s a fixed amount of money that is paid on a regular schedule (monthly, biweekly) to all citizens without being means-tested or having to fulfill work requirements.
In the case of the Yang’s plan, he is talking about 1000$ per citizen. Which comes astonishingly close to the 13000$ per year and citizen that Charles Murray proposed in his book “In our hands”. I got the 2016 edition, but the book as been around since 2006. I don’t know if the numbers were the same back even then. Murray wants people of the age of 21 to get it, Yang is aiming for 18 and above.
If we do the math, we’ll quickly see that we are talking about a volume of trillions of $ per year in the USA. That’s whole lot of dough!
So how are you going to fund it? Interestingly enough, most proposals I have seen assume that to be an easy question; they either want to increase tax revenue or stop other programs or use a combination of both.
Murray for example claims that “[t]he crossover year in which the UBI [his version] became cheaper than the existing system was 2009” [MURRAY, p.11].
He assumes that the existing tax code stays in place and eliminates “Social Security, Medicare, Medicaid, welfare programs, social services programs, agricultural subsidies, and corporate welfare.” [MURRAY, p.10].
So every payment is bundled up in the UBI. You also get a universal passport (“This passport also establishes eligibility for the UBI“), a bank account (“No back account, no grant.“) and catastrophic health insurance (for which you have to pay 3000$ out of your 13000$). Immigrants don’t enjoy a UBI until they get naturalized.
Yang proposes the introduction of a VAT at 10%. “A VAT will become more and more important as technology improves because you cannot collect income tax from robots or software.” [YANG]
He wants to reduce payments on “food stamps, disability and the like” and hopes that “[w]e would save $100 – 200+ billion as people would be able to take better care of themselves and avoid the emergency room, jail, and the street and would generally be more functional.” [YANG] Additional revenue would be generated “[b]y removing the Social Security cap, implementing a financial transactions tax, and ending the favorable tax treatment for capital gains/carried interest[.]” [YANG]
[STRAUBHAAR] proposes a flat tax of 50% on all forms of incomes levied at the source; including all interests, income from rents, royalties, robots etc. Somebody gets paid for having these robots and apps produce stuff and services. He’s going to tax that.
This would result in a progressive tax regime, as this table illustrates:
The progressiveness of this scheme is actually something he emphasizes and contrasts to indirect taxes, like a VAT in Yang’s proposal; if you are poor, you will spend a huge percentage of your income every month, which will be taxed. The result could be highly regressive.
The general arguments for why a UBI is preferable are rather similar.
Elimination of the “welfare cliff“. Current social benefits reduce the incentive to work, because one additional dollar in income can cost you a multiple of that in lost benefits. The implicit tax rate poor people face is often staggeringly high.
Reduction of bureaucracy. Sending out a check to a list of people is easier than collecting information about their income first.
Increased self-esteem. People, who have an alternative stream of income are less likely to tolerate abusive work place conditions. The stigma of having to live with government transfers is reduced. People have “their own money” and are given agency.
Increased entrepreneurship. The UBI provides for the basic needs during the early phases of a company and is a safety net, if the business fails (which it does 90%+ of the time. Keep at learning! The next might work out for you!)
Improved mental health. Worrying is not healthy. Having less sorrows, because you have dependable income stream is projected to help with stress.
Improved labor market efficiency. Because people keep their health insurance and a dependable income stream, they will be less likely sticking with a secure dead-end job.
I’d like to add that a carbon dividend, for which I argued previously, could be easily integrated into this system. Even more so, if the infrastructure for a VAT, like in Yang’s proposal is implemented. The carbon tax would just be a second position on the VAT statements. In any case, the dividends could be added as additional item on the UBI transfers.
There is a whole list of people who support(ed) a UBI, Milton Friedman favored the “Negative Income Tax”, as did Rawls, Elon Musk says it’s “necessary”, Mark Zuckerberg, Richard Branson, and Sam Altman seem also supportive of the idea.
They probably don’t imagine themselves to be the cruel and bored elite of the Mad Max scenario, but I can’t read their minds.
Criticisms
In some small scale experiments, a few of the most frequently heard arguments against a UBI, that it induces drug usage and general laziness, could not be substantiated. But then again, it was necessarily small scale and often in countries that aren’t really representative of industrialized countries.
So my current thinking on that topic is well expressed by Murray
The fact that no able-bodied person needs to live in poverty doesn’t mean that no one will live in poverty. Some people behave in ways that ensure they will live in squalor, will not have enough money to buy food, or will be evicted for not paying the rent. They may drink away their money or gamble it away. Some people will be feckless under any systems. The UBI ends nearly all involuntary poverty — the kind that exists when people have done the ordinary things right and are still poor. [MURRAY, p.42]
The UBI is a systematic approach to give all people the means for their needs. We can’t really control their choices after that. UBI is not a panacea.
There are of course people, who think that the UBI should be way, way higher. Why not go with 1500$ per person? Or 2500$? The Swiss people rejected the latter proposal in a plebiscite not too long ago. But of course there is nothing intrinsically special about that amount of money. So setting the amount and especially the progression of the payments might become a point of fierce contention. Although that might be healthy for society at large.
And of course there are critics to the idea, that think it’s way too expensive. “UBI is a flawed idea, not least because it would be prohibitively expensive. […] Though UBI makes for a good slogan, it is a poorly designed policy” says Daron Acemoglu, a reputable economist, in a proper trouncing of the UBI, recently. “Besides, a more sensible policy is already on offer: a negative income tax, or what is sometimes called ‘guaranteed basic income’. Rather than giving everyone $1,000 per month, a guaranteed-income programme would offer transfers only to individuals whose monthly income is below $1,000, thereby coming in at a mere fraction of a UBI’s cost.” Ouch! But he doesn’t stop there: “As such, UBI proposals have all the hallmarks of the “bread and circuses” used by the Roman and Byzantine empires – handouts to defuse discontent and mollify the masses, rather than providing them with economic opportunities and political agency.” Well, that is probably as stingy as it is true.
Reflecting on it, it have to admit that a part of UBI’s appeal to me is exactly that it plasters over a problem: the loss of economic value of a whole lot of people, by alleviating a symptom: no income.
I have no idea, how to cure the former, so I’d probably still opt for the latter. It’s analogous to home invasions: the “real problem” might well be poverty, drug abuse, testosterone, etc. Doors and locks still help.
If we find a way to keep most people competitive with robots and increasingly sophisticated automation, I’ll opt for that. If we don’t know how to do that, spreading money and Netflix Premium accounts around to “mollify” the masses might be the best available option.
Alternatives
Anyway, the case for a UBI might be not as clear cut, as I’d like it to be. It’s conceptual simplicity and the promise of decluttering the byzantine nightmare of regulations, systems and programs seems really appealing to me. (The welfare state is just horrible at UX!) But it seems more than likely that societies will have to make risk-aware intermediate steps to get there. Which sound like the position of Bill Gates on the topic.
Maybe, a GBI could be less risky start. The robots aren’t here yet, so let’s learn what we can, while we are not in full crisis mode and can iterate with less risks.
The question becomes: are there more alternative ideas to ensure the social minimum that Rawls’ conception of justice demands? And of course there are. People are just amazing a producing ideas!
Universal Basic Services
A rather boring, if I might say so, conception is the Universal Basic Services (UBS) idea, spearheaded, as it seems, by Anna Coote and Andrew Percy, who co-authored “The Case for Universal Basic Services“. It defined as “the provision of sufficient free public services, as can be afforded from a reasonable tax on incomes, to enable every citizen’s safety, opportunity, and participation.” [UBS]
When I say it’s boring, I don’t mean to say it’s not worth considering. But it just seems like an extrapolation of current European welfare concepts mixed with a huge dose of old-fashioned skepticism towards capitalism and markets in general. But that might actually be it’s strongest asset: it seems entirely practical, “traditional” and affordable. (despite repeated claims about the proposal’s radicalism by the authors) For a package of universal free “basic” housing, free food, free local transport, free TV licence and a free‘”basic” communications package including mobile phone and broadband internet, they estimate for the UK a cost of £42 billion or about 2.3% of GDP. Healthcare and education are already provided by the state and don’t show up as cost of this proposal.
Scott Adams recently called a form of UBI “inevitable”, but proposed the usage of crypto-currencies for the UBI, which would only be redeemable for essentials. I think that’s structurally closer to the UBS approach, as society would need to decide on what’s essential and make it accessible to all instead of distributing money for whatever purposes the recipients deem appropriate. And that draws of course fierce criticism about the creation of a “paternalistic state”.
Maybe a plebiscite could help to determine, what society at large and not political parties deem necessary? Or maybe deliberative polls? If this conception is really as efficient as is claimed, this might be worth a try. If the UK opted for that route, could the EU and US assist with statisticians? Or how about testing it in one or multiple cities or counties? Can we learn something from free public transport in Luxembourg? This conceptions seems entirely A/B testable.
Universal Basic Assets
Another alternative is called Universal Basic Stake(holding) or Universal Basic Assets (UBA).
The idea is to give every citizen “skin in the game”, by giving every citizen a share in the assets of a society. This could, but does not necessarily has to mean, that the state buys large portions of existing companies’ stocks, engages in venture or even angel investments to create new companies or builds, operates and/or finances housing, energy and transportation infrastructure and so on. We could also use so called “golden shares” on patents and technologies that get created based on public research, like Mariana Mazzucato argues for in her book “The Entrepreneurial State”, that create equity for the state.
The return on the investments belongs to all citizens equally. It’s like UBI, but with an extra twist: instead of wrestling with the owners of the robots, algorithms and data to get them to pay “their fair share”, make the people (at least partial) owner of those assets.
This sounds an awful lot like socialism, but I actually don’t know how close this comes to socialists idea(l)s. It seems more inline with what market-oriented critics of the typical pension schemes have been advocating for for decades: getting people to opt for an equity-based retirement option. See for example **Laurence Kotlikoff’**s “Purple Social Security Plan“. Except that the payments start rolling in at 20ish instead of 70ish.
Few people would accuse Singapore, which gets consistently ranked as one of the most economically free countries in the world, of being socialist. But Singapore has the Central Provident Fund, which is a compulsory comprehensive savings and pension plan, aiming at retirement, healthcare and housing needs. In Singapore, your savings belong to you, but you are required by law to save a certain amount that gets managed by the CPF. You can use it to pay for certain medical expenditures, which is supposed to make you “shop for alternatives” and keep costs reasonably low (which seems to work great, by the way, if we can believe William Haseltine‘s “Affordable Execellence”), but you can also use it to borrow against, if you want to buy an apartment.
So, would a scheme like that, where everybody pays in a percentage of income, but, in difference to Singapore’s model, gets out a per capita share, be workable?
I don’t know.
What problems would we get in a world, where assets are spread more widely? One is probably that the authority to make decision is thinly spread out. I am also unclear about the process, with which we would send decision-makers on behalf of the state to the stakeholder meetings. Or how we could hold them accountable. I am pretty sure we will need a lot of random samples to eliminate shenanigans. Then there seem to be a few problems with too much centralized control that politicians could muster, but maybe that could be combated by some of the ideas, I presented in a previous article. Is it just a job for the distribution branch of the government?
There are a whole lot of questions I have for this conception of ensuring the social minimum, but it seems to be even more “pure” then the UBI to me. It’s not only concerned with ensuring a share of the flow of the wealth of society goes to all citizens, it also aims at given broad access to the stock of wealth.
“Society” could be the owner of the UBS it administers via competing, private companies to herself. Are UBI, UBS and UBA even distinct concepts or do they just put different emphasis on the solutions?
Anyhow, results are more important than my particular taste for ideas, so we need to test the ideas. Any suggestions, how we could scale that down small enough to test it with reasonable risk?
Final thoughts
My deliberately chosen interpretation of justice demands a social minimum.
“There is with reason strong objection to the competitive determination of total income, since this ignores the claims of need and an appropriate standard of life.” [RAWLS, p. 277].
The systems we have invented to get close to this are coming under increasing strain. An aging population, huge levels of national debt and the looming “threat” of automation make it more important than ever to think about new ways to ensure the social minimum. I think it’s unproductive to let people suffer the results of a bad lot in the lottery of nature, technology and society. But a goal, even a lofty one, is not sufficient. We need a system to get there. And especially a system that works with real human beings, which almost never behave like rational decision-theory would have them behave. Rather, our biases seem to be the basis of our evolved humanOS. We have to deal with that. There are multiple proposals for these systems and we still have an awful lot to try, before we know what works and what doesn’t.
Let’s get going!
Sources:
[MURRAY] MURRAY, Charles. In our hands: A plan to replace the welfare state. Rowman & Littlefield, 2016.
[RAWLS] RAWLS, John. A Theory of Justice (Oxford Paperbacks 301 301). Harvard University Press. Kindle-Version
[STRAUBHAAR] STRAUBHAAR, Thomas. Radikal gerecht: Wie das bedingungslose Grundeinkommen den Sozialstaat revolutioniert. Edition Körber, 2017.
[UBS] from:
https://universalbasicservices.org/
. Accessed 2020-05-09
[YANG] from: https://www.yang2020.com/what-is-freedom-dividend-faq/. Accessed 2020-05-09