The Virus, the Robot and The Transformation of Capitalism

The Virus

Tyson's, one of the largest processors and distributors of chicken and meat products in the world, has been automating their production facilities for some time, but the process gained a whole new urgency last April when a worker tested positive for Covid-19. Subsequent inspections revealed that the virus had made its way into the meat products being prepared, forcing a hasty recall and temporary closures as they disinfected the facilities. Ultimately, more than 17,000 workers were infected with the virus and 93 died, one of the worst single company outbreaks in the United States.

Covid-19 has upended the political landscape, has idled more than 30 million workers and has been responsible for more than 130,000 deaths in the last five months in the US alone. It has forced a heated and vocal debate about health care, the wearing of masks, and has set in motion a recession that likely will play out for years. Hopes that Covid-19 would eventually go away have proved in vain; the closest comparison to how it's playing out is the Spanish Flu outbreak of 1918, which ultimately ended up killing millions worldwide and was still influencing society years later.

Tysons responded by investing heavily in robotics, significantly reducing the workforce of butchers, processors, and inspectors in the process. The need was obvious - human workers had become vectors of infection for their products, and so long as Covid-19 is a factor, having humans in the loop would prove costly. Where things get problematic is afterward.

It's possible that the pandemic may prove to be a once in a lifetime event, but even as this particular virus recedes in Asia and Europe, it is raging out of control in the United States. The world has managed (mostly) to dodge pandemics for the last several decades, but the reality is that such pandemics are becoming harder to predict and contain. Environmental factors, including massive urbanization in China, a warming climate, deforestation in Brazil, global air travel, and so forth are providing new environments conducive to the creation of pathogens.

The threat of genetically engineered pathogens is also becoming more real, as low-cost genetic sequencers and increasingly precise CRISPR toolkits raise the potential for "tinkering", potentially with malicious intent. Covid-19 was probably not genetically engineered, but it's all too easy to see it being modified as a general delivery platform for viruses that were.

The Robot

What that means is that there's a reckoning that we as a society have been putting off, perhaps for too long. Automation is not new. The Industrial Age was built on automation, on the creation of increasingly complex machines. What has changed is that up until recently, those machines were incapable of making more than the most rudimentary of decisions.

Today, the typical "robot" is able to see, to hear, to taste and smell, in most cases to a degree that is far superior to our own primary senses. It is able to learn, in the sense that it not only can follow an algorithm but that it can generate its own algorithms based upon training ... and what's more, can use reinforcement learning to adapt to unforeseen circumstances. These robots are specialized AIs - they can't necessarily reason to the level that humans can, but they are getting much better at reasoning about specific tasks at a level comparable to a moderately skilled human specialist.

To illustrate this, consider the aforementioned meat workers. A new butcher will typically take a few months to learn how to efficiently strip a carcass, partially because there's knowledge involved and partially because that butcher needs to develop the muscle memory to do that task efficiently. It may take a couple of years before they achieve a decent level of mastery, and typically they work most parts of a processing run to gain an overall understanding.

An AI-based robot is not good at all things, and even with machine-learning-based training, it can take a while to develop a good model to be more than mediocre even at one task. The difference, however, between humans and robots is that once you develop that model, you can port the same model to different machines at a minimal cost. You end up with robots with subpar skills, but with the potential to improve those skills over time, that can operate 24/7, and that amortize at far below a human worker.

Up until now, these kinds of considerations were balanced with the issue that it also effectively destroyed jobs, and as such there were frequently political concerns that also had to be allayed. Getting tax breaks from local governments by promising jobs has become a popular technique to both get funding and clearing the necessary permits, but the case becomes harder to make when you're basically putting in robotic assembly line systems. Yes, you still need robotics experts to help to train the models and repair the bots, but you need far fewer of them, and that tends to make the political argument harder.

What Covid-19 has done, however, is to provide a counter-argument: you're potentially jeopardizing people's lives if you have too many human workers in the mix. There is, unfortunately, enough to this that it may very well become the rallying cry for corporations to automate more and more of their existing infrastructure while reaping significant savings in terms of labor costs.

There is, at the heart of capitalism, a contradiction, one that we are now running up against. To participate in the economy, you need money. In the industrial twentieth century, that money was made through two means - you either earned money by exchanging your labor for some other person's wealth, or you made money by providing capital to others with a promise of a return on the investment.

With drones, robotics, the Internet, AI and other automation (and especially as 3D printing starts ramping up on metals) you have less need for extractive workers (miners, riggers, fitters, welders), less need for construction workers (plumbers, electricians, contractors), less need for agricultural workers (pickers, butchers, bakers, farmers), less need for security (police, firefighters), less need for retail people, in short, less need for anyone who isn't an investor (and yes, this includes managers and salespeople here). Of course, this also means that those same people are no longer participating in the economy. This means they are no longer buying what is being produced by anyone.

Those who have the bulk of the wealth may be able to skate for a few years on what they have built up, but eventually, even that will not prove enough, simply because the wealthy 1% only accounts for perhaps 5% of overall spending, and even the wealthy 10% account for perhaps 15%. As the average consumer's spending dries up, so too does the economy.

 

Automation's Role In Capturing Productivity

I want to make a distinction here between what is a long term trend and the current recession, though I suspect that the latter will only exacerbate the former. Even given the current state of specialized artificial intelligence (which can arguably be described as fairly primitive) there are very few jobs that cannot be done by a machine. Tellingly, most of them are physically active jobs. It's fairly easy to replace an insurance adjuster, for instance, but is considerably more difficult to replace a fry cook. The question ultimately is whether or not the cost of replacing that fry cook with a mechanized version is worth the effort.

I have had, over the course of my life, a number of jobs that were ultimately replaced in one fashion or another, first by work that was automated, then by work that was outsourced before it became automated. There is a pattern where jobs first become augmented by AI, but in the process of developing this augmentation, enough intelligence is baked in that the AI is eventually capable of creating new content directly, eliminating larger and larger swathes of human workers in the process.

Robots (or more properly AIs) will not replace all workers, but they have the potential to ultimately replace a significant percentage of the total workforce. We need to acknowledge this as a reality, one that is happening today. One large commercial bank, as one example, recently announced that they are cutting tens of thousands of workers in response to Covid-19, the first likely of many such announcements in the coming months. At the same time, that particular bank (and most others) have been automating most of its transactions, including mortgage and business loan applications, depositing of checks, and setting up accounts, so that these can be done directly by customers.

The question that emerges here is a fundamental one: eventually, some means will be found to mitigate the effects of the Coronavirus. When that happens, will those banks start rehiring new tellers, loan officers, or even build satellite banks? Probably not, because there will be insufficient need to justify that hiring. Note that this doesn't mean there will be no need, only that the cost of providing human customer service will not be deemed necessary for any but "high value" clients (and that high-end customer service will likely be the last cut).

What this means in the intermediate-term (over the next three to five years) is that we will likely be facing a period that was similar to the from 2010 to 2015, where employment growth seemed slow and wages remained stagnant, but this time it will likely be worse. AI systems may be inferior, but they are also far cheaper, especially given the likelihood that such solutions are going to face a significant discount as (increasingly independent) software development teams take what they can get.

Companies will likely find that while they require fewer workers, they will also find that they are selling fewer products or services because only the wealthiest have money. We were heading in that direction anyway - in the field of entertainment, for instance, there has been a significant growth (that Covid-19 likely hasn't affected much at all) in the arena of specialized or bespoke videos produced only for private customers, even as movie theaters are already closing.

This will eventually result in economic stagnation and possibly a long drawn out depression in the US in particular. There's been a long-standing belief that automation will bring with it new jobs. For a while, that was true, partially because automation required skilled programmers, designers, managers, and their associated support personnel to build applications. and the applications that they created tended to in turn facilitate (augment) people's creativity and productivity.

Now, however, there are few areas where there aren't already very mature tools and products already crowding one another out, along with a huge amount of open-source technology that has gone from barely useful to good enough. The world doesn't need another office suite, another health app, another point of purchase platform. We're drowning in applications, and as such, the value of such applications are dropping quickly.

AI only accelerates that trend, because with it all you need is enough data to analyze patterns and create neural networks. You don't even need to understand how such networks work, because we've already automated most of those processes that as well. Yes, there may be a need for a few PhDs to develop new algorithms (not that there aren't already more than most data scientists are aware of) but those are highly specialized jobs for a comparatively tiny percentage of the overall population.

Thoughts on Bots

I've debated about whether to put this section in this article, but I think it is worth stating. Automation ... AI ... changes everything. It changes notions about authority, responsibility and the boundaries of freedom. The Internet and social media (which are all AI constructs in the broadest sense) amplify voices, but the problem is that it can also make many voices louder than they actually are, or legitimately should be. We, individually and collectively, are learning how to filter out the worst of it, but it's an arduous process, made worse by the fact that often the voices that are loudest are typically the ones that are intended to steer people away from their best interests.

The technology stack behind artificial intelligence includes the ability to create shadow people, bots in the vernacular. A bot is an avatar (a virtual representation) of a programmable agent. Bots can pass the Turing test, meaning that it is not always obvious that the person on the other side of the screen is human or not. This is partially due to the fact that the technology of AI, while still basic, has been improving dramatically in the last few years, but it also has to do with the fact that most people are generally not trained for deep analytic thinking, and in general their interactions with most other people are at a fairly superficial level anyway.

It is this last point that needs to be stressed. Bots do not need to necessarily be that intelligent. They simply need to be intelligent enough so that people believe, upon first interaction, that they are dealing with another person. Humans invest emotion into their connections, we love it when others agree with us, get indignant when others get upset, and take a secret delight when someone we despise gets their comeuppance.

Skilled bot-writers understand this, and use it to manipulate emotions, bypassing logic and causing us to suspend our skepticism and disbelief. This is one of the reasons that bots can be so dangerous. We have close to a hundred years' worth of understanding about how marketing, propaganda, and advertising works, which ultimately comes down to determining how to "push people's buttons" in order to steer them towards taking, or not taking, certain actions.

As bots become more human-seeming, we become more vulnerable to manipulation because our trust levels tend to be higher in one-on-one communication than it does with broadcast (or via literary channels), especially when we perceive someone to be friendly without being manipulative. In a sense, we are reaching a stage where bots are now artificially authentic, and that is, at its core, fundamentally worrisome, because one of the major uses that companies have for workers is to communicate with other people - to sell, to teach, to influence, to justify. When those kinds of jobs succumb to automation (which is now happening rapidly) then it is very likely that the economy will collapse, because there will no longer be anyone able to buy what's being sold.

Several years ago, during the throes of the 2008 recession, I remember one commentator on an economics-oriented talking heads show bemoaning about the fact that "consumers weren't yet willing to open their wallets for business, that they were reluctant to spend". I wanted to throw my shoe at the screen at the time. This framing perpetuated the belief that people had money, they just needed to give it up, rather than facing the reality that people had nothing to buy with. The desire may have been there to buy, the goods were there, but the fiction that consumers and workers are not connected was fully exposed there.

In a way, we are facing multiple versions of the Tragedy of the Commons, where everyone's desire to feed their own cows ultimately destroys the commons that they were feeding from. In this case, the Commons can be thought of as the collective money that consumers have, and every entrepreneur's desire to make money off that particular group. The problem is that without some way of replenishment, that resource will eventually drain dry. In the context of the pandemic, that "eventually" will likely be very soon indeed.

Facing Hard Choices

When a large enough percentage of a society's middle class becomes functionally unemployed, you get massive unrest, and typically that results in governments falling. However, what we're facing is a systemic problem, one that likely transcends political parties, though how each such party attempts to solve it will more than likely determine their viability in the next decade.

There are several alternatives moving forward that can potentially salvage the situation, but there will be a cost on some group or another that will have to be borne.

Do Nothing

This is the scenario that exists today. Economic inequality increases to a point where the velocity of money collapses altogether, starvation becomes wide spread while the wealthiest flee the country in pursuit of "safer ground" even as the US enters into a full bore authoritarian state. Eventually, the wealthiest regions of the US will become de facto (and then de jure) countries, while other regions become increasingly exploitative of their labor and resources, to the point where these regions eventually play out their reserves on both fronts and become failed states. The US withdraws more and more of its presence from the world stage, losing both opportunities for export and revenue that has come from the protection that the military provided. Innovation slows and moves increasingly away from the US, even as the educational system collapses. This won't happen overnight, but there are signs that some of this is happening now.

Tax the Proceeds of Automation

No one likes taxes. Most people believe that the tax code is unfair, but as to whom the tax code favors, opinions vary considerably. I would argue, however, that there are several compelling reasons why the proceeds from automation absolutely must be taxed. Most of these arguments ultimately come down to the question of who deserves the benefits from automation - the creator, the investor, the customer, or the citizen. Ultimately, and historically, the investor has benefited far and away more than any of the others, and the question ultimately comes down to whether such benefits were deserved. I'd contend the answer is usually no, with caveats.

• Automation typically is based upon products and platforms that were produced by collective efforts over the course of decades. Most of the core algorithms in use today were developed primarily by people working in academia or research facilities, and typically distributed at little or no cost. Linux, the most widely deployed operating system in the world, is open source. The Internet was based on the work of Tim Berners-Lee, who famously felt that it should not have been encumbered with a commercial license. Most of the ideas that laid the foundation for Neural Networks came out of work from Marvin Minsky's MIT Laboratory for Artificial Intelligence. Python and Javascript both were open source languages.
• A significant amount of the work of developing applications on top of these typically involves the use of software professionals who either funded themselves through school to gain skills or were perceived by those schools to be worth funding for their potential to make a difference. Companies pay for the products that they produce with this training, but typically the ones who benefit most from this are not the software developers (many of whom are frequently let go during economic slowdowns, with little to show for their work in often highly inflated markets) but investors and senior salespeople (including CEOs).
• Most investments made into companies are not made during early seed-rounds. Instead, they are made only after a particular company has proven itself and is making a significant profit, at which point, they are considered safe investments (they produced dividends reliably). This is where wealth generation kicks in, and it actually benefits no one but the trader and the senior stockholders of the companies in question. Indeed, it can be argued that once a company goes public, it is in that company's best interests to buy up competitors that it can't force out of the market, usually only benefitting the small percentage of people who were vested in that company, reducing potential innovation and frequently leaving those who were underpaid for their skills scrambling to find new work.
• A second source of inequity comes from the broad use of contractors, frequently insourced from countries with lower wage expectations and costs of living, who generally do not participate at all in benefits beyond the (often shaky) prospect of continued work. This usually serves to reduce wages for technologists and technology support, sometimes dramatically, with most of the profits then going through multiple layers of contracting agencies to be aggregated by large general contractors.
• Still another potential revenue source is a trading transactional fee, basically, a sales tax that gets placed on purely electronic transfers. This has perhaps the biggest applicability in areas such as high-speed financial transactions such as front-running stock purchases (something that's dubious in its own right). The amount would be comparatively small for the bulk of transactions, but especially in the arena high-frequency trading, the amounts involved would add up quickly (and would have an additional benefit in that it reduces the potential for volatility in the market).
• One additional idea that is increasingly making sense - it should be possible, when filing taxes, to specify how a portion of tax receipts per individual get allocated. Even if only a portion of proceeds fell under this domain (think of it as discretionary funds) it basically provides a good benchmark to determine what the priority of all Americans are, not just simply the wealthiest, and it reduces the role of lobbyists in the political process.

Universal Basic Income, Paid Education, and Universal Healthcare

Here's a question that is not asked very often. What would happen if businesses could reduce their per employee costs by $40,000 a year? This is roughly the cost that each employee pays in order to live close to work and to have healthcare, plus amortization on their education debts. It's $40,000 that those same employees would have in their pockets to spend on the goods and services that businesses are in the business of selling. It is also, roughly, the amount that most businesses are now pocketing as "productivity gains" due to automation, in effect taking advantage of that automation to need fewer workers, but also creating far fewer customers.

Here's another question: Why should getting an education be so expensive? It's expensive in part because companies have traditionally benefited from having someone else do the hard, expensive, time-consuming job of training people. However, it also means that the number of people who have the specialized skills that the company needs will also be limited, which drives up the costs of those skillsets, often beyond what many companies can afford.

The employment system in the United States evolved dramatically during the 1920s through the 1940s. Most of the work at that time only required a high school education. You gained expertise in a company by experience, gaining a breadth of knowledge about the company and what it produced, to the extent that someone could start out on "the floor" and still potentially make it all the way to the president's office.

Automation, especially computerized automation, changed all that. In many fields, there are no pathways to advancement within the same organization. The average recession now comes roughly once every eight years and the chances that you will be laid off during those recessions is now more than 50%. If you are a contractor, your chance of getting on full time is as low as its ever been. If you are a part-time worker, your chances are even worse.

Churn has become the norm, and it is becoming commonplace for people to be out of work for roughly six weeks every year. Unemployed people tend to fall out of the system in greater numbers as well - the worst thing that can happen on your resume is for you to be unemployed for longer than a month because, at that point, there becomes a question of whether you are in fact employable. When you are unemployed, it can take a while to get assistance, and in many states, the odds of getting any assistance are disturbingly low. Without money, you lose health care, you cannot pay for increasingly expensive schools, and you end up in an endless spiral where your career momentum stalls, and you're forced into finding any job at significantly reduced wages - assuming you don't end up on the streets.

So, suppose that you taxed most of the gains that companies made by using automation to reduce headcount, and made a certain percentage of that, say $3600 a month, available as a univeral basic income. Suppose that you taxed dividends that came from using labor that was educated by money the laborers themselves paid, and used it to provide free tuition and fees so that there was an adequate supply of skilled labor. Suppose that companies stopped paying exorbitant and growing premiums to insurance companies and instead paid into a single universal pool.

Given all of this, you would think that it would make clear sense to make some kind of UBI available. There are, however, several things standing in the way of that happening. The first is, of course, entrenched interests that benefit from the status quo, from staffing agencies that take a cut of contract wages (usually while providing only a trivial amount of value to the employee) to businesses that use employment arbitrage rates to generate profits to health insurance companies that would be forced into working with a much smaller pool of (likely older) applicants for supplemental insurance purposes.

Additionally, there's a concept that gets bandied about frequently in conservative circles: moral hazard. The idea here is that if people were no longer so desperate to work, then they wouldn't work, and there wouldn't be enough labor to fill the needs of companies. Moral hazard is also frequently framed in terms of laziness - that without incentive to work, people would become lazy, and would fail to be productive in society. This particular argument frames such indolence or idleness as being a gateway to evil actions. It's also a tautology, as there's a fundamental question about what "productive" actually means.

In a society where labor, hard physical labor, is needed, there may be some validity to this argument. However, it's worth noting that most labor today is focused primarily upon making investors wealthier. Automation is frequently sold as labor-saving, yet the labor being saved is very seldom passed back to the worker - it's transferred to investors as dividends.

Most people, given free time and the luxury of financial security, may actually take a vacation or be indolent for a while, but boredom if nothing else usually motivates people to do something: to create art, write novels or essays or poems, create videos, build software, solve mathematical problems, fix cars, teach, work with others to create music, help people in need, save species from disappearing, reclaim damaged habitats, raise children, grow gardens, you know ... live life.

Summary

The pandemic may, ironically enough, be the opportunity to question the directions that we have taken. This is not to say that capitalism itself is wrong, but there is growing evidence that it has metastasized into something toxic, something that benefits a very small number of people largely by robbing other people of their livelihoods and lives.

We should stop being worried about losing our jobs to AI and start asking ourselves why we shouldn't be benefitting from several generations of people standing on one another's shoulders to create that AI in the first place. Who benefits from all that labor-saving code, art and ideas? This ultimately is what open source and open licenses were all about in the first place. It is time for us to acknowledge this reality, to recognize that when you have the means to reduce unnecessary productivity, everyone should benefit.