Mid-Term Effects of AI on G7 Labor Markets
Or, an Entrepreneur’s Guide to the 21st Century
Wednesday, November 1, 2017
I. Introduction
Automation and Labor: A Brief History
I recently read an article in TIME magazine which went: “What worries many job experts more is that automation may prevent the economy from creating enough new jobs… Throughout industry, the trend has been to bigger production with a smaller work force… Many of the losses in factory jobs have been countered by an increase in the service industries or in office jobs. But automation is beginning to move in and eliminate office jobs too.”
The article sounded like so many others one comes across in the business section. But there’s a catch — it was published in February, 1961. Indeed, such was the furor surrounding automation at the time, President Johnson empaneled a “Blue-Ribbon National Commission on Technology, Automation, and Economic Progress” to confront the looming prospect of mass unemployment. (The panel ultimately concluded that consumer demand, not technological progress, was the primary determinant of job growth, and hence the hysteria was unwarranted. Or as it was put it the panel’s white paper, “technology eliminates jobs, not work”.)
Mass media of the 1960s can perhaps be forgiven for fanning the flames of automation anxiety. After all, the quantum leaps being made in computing wiped out legions of secretaries, CNC machines out of MIT were displacing machinists, and automated assembly lines from Yokohama to Youngstown were putting able-bodied workers out of jobs. On the agricultural front, self-propelled combine harvesters sent farmworkers packing, while new pesticides obviated the need for many thousands of human weeders. In other words, there was far more visible, even unprecedented, evidence of technological unemployment then than there is today.
To be fair, automation anxiety has existed in one form or another since the dawn of recorded human history. Scribes and stationers were irate after Gutenberg commercialized his printing press. Queen Elizabeth I, called being upon to adjudicate a 1598 patent dispute concerning an automated knitting machine, replied to the inventor “Consider thou what the invention could do to my poor subjects. It would assuredly bring them to ruin by depriving them of employment, thus making them beggars.” Her caution failed to forestall the luddite uprising 213 years thence, which centered around v3.0 of the very same knitting machine.
Fascinatingly, the employment debate today closely mirrors the mood if not the headlines of that many centuries past. The crucial difference is that, this time around, the press is not fretting advanced machines so much as advanced code — and in particular the code driving artificial intelligence (AI) applications. AI applications are so interesting because they promise to transform many tasks — from the way diseases are diagnosed to the way doughnuts are delivered — and, in the process, cut out the human intermediary. The pertinent question regarding AI thus becomes, is it different in degree or in kind from previous technological developments that affected technological unemployment? One would hope that AI is the former; that it’s merely a continuation of a millennia-long megatrend where technology advances, workers adjust, and we all go on our merry way. However, in the case that AI is the latter, humanity will have a situation on its hands — one where we become slaves to a robot master race, or alternately, lead lives of blissful leisure, depending on whom one asks.
II. Predictions, Prophesies, Prognosis
My belief is that AI-related technology represents the first sort of change. The upshot ought to be that no major change is to be expected. We all go on our merry way, right? Sadly, I would argue, there is a caveat that must be attached to even development-by-accumulation type change (as opposed to paradigm shift type change, analogizing from Kuhn’s framework). Namely, public institutions must be capable of responding in a rational, self-interested way to such change. While Gutenberg’s Holy Roman Empire, Queen Elizabeth I’s England, and LBJ’s America — to borrow from my previous examples — were far from optimal in terms of public policy, they were undoubtedly self-interested empires, and indeed consistently instituted policy that would further their wealth and stability. One need look no further than Queen Elizabeth I’s promotion of tolerance and trade (enabling Shakespeare to write and Francis Drake to circumnavigate the globe), or LBJ’s radical Great Society legislation to see this borne out.
The question as to whether the US, or even Britain or the EU, is — in 2017 and beyond — capable of consistently issuing rational, self-interested policy is an open one. This is not so much a swipe at the White House or Brexiteers as a commentary on the inertia of the state in the 21st Century. That is to say, there are so many citizens and so many programs to administer, that those in charge — however well-intentioned, however well-informed — scarcely stand a chance of comprehending what would constitute an optimal program, much less delivering it to their citizens. It’s a case of complexity overload. There’s a good chance for this reason that there is no longer a way for large countries to govern optimally on a regular basis — at least not without help from an AI that could compute optimal public policy.
One might come to this conclusion from a more casual direction. To wit, those in the executive and legislative branches of the US government are so incompetent, bribed, or otherwise ill-equipped to lead, that impactful and productive legislation is all but impossible going forward. This is the argument I favor. It is bolstered by the performance of Presidents Obama as well as Trump. In the first case, setting aside politics, America had a highly disciplined, intelligent, and energetic Commander in Chief who installed in office many thousands who reflected those characteristics. In spite of what might be considered a close-to-best-case scenario, little legislation of note was passed into law, and the nation’s gravest structural problems remained unaddressed. In the second case — what might be considered a close-to-worst-case scenario — no legislation of note has passed, and in the unlikely event it does, it will assuredly be rubbish of the worst order. One can expect structural problems to accumulate and fester, with all that entails: talent fleeing, corporations stonewalling, tax receipts declining, and municipalities responding to such stimulus as they always have — by barricading themselves into self-reliant fiefdoms.
(For precedents here, look to the city-states of ancient Greece, the antichi stati of the renaissance-era Italian peninsula, or even cities of the colonial Americas — where, post-succession, each state was forced to issue its own currency.)
These fiefdoms will compete amongst themselves in an increasingly winner-takes-all state of play. Admittedly, this is hardly a break from the status quo, where New York owns finance and Los Angeles owns entertainment, so much as an accelerated continuation. Regardless of whether China assumes the role of leading superpower, many urbanists believe that such balkanization is already guaranteed throughout the US and Europe. This, in short, is why it’s impossible to say that AI will have a singular effect on G7 labor markets.
To the contrary, one ought to expect AI-driven economic outcomes to highly diverge depending on a city’s specific mix of industry, education, wealth, proximity to natural resources, etc. Incidentally, the most important determinant of this outcome is population size: cities with large populations are predicted to fare better, whereas cities with fewer people are predicted to take the brunt. (This prediction is made in study after study after study.) Economies of scale, virtuous cycles, and network effects — not a town’s mettle or its livability — will allocate the rewards and mete out the punishment of AI’s rise.
In sum, there is a disjointed prognosis for the American economy as it grapples with the effects of accelerating AI-driven automation. Large, rich cities over the coming decades should get larger and richer. They will be the happy hosts of the supra-API class poised to dominate 21st Century labor markets. Midsize and small cities, on the other hand, should ceteris paribus expect to see continued decline. Many of their factories will shutter or relocate. Their counties will demand more public assistance, as residents are increasingly afflicted by joblessness, depression, and drug addiction. As to which politicians are elected to clean up this mess, one can only tremble in fear. The rabble entering national office will further undermine the wellbeing of the ‘losing’ cities, advancing America’s balkanization.
III. Business Model Opportunities: A Speculative Overview
What are well-meaning millennials to do in response to the scenario described above? One answer is undoubtedly to flock to one of the cities that has cashed a winning ticket in the urbanism lottery. There, one is free to become a modern-day Robert Moses or Richard Daley, pursuing the best interest of one’s neighbors in a context amenable to control. Alternatively, one may hold out hope that the federal government has a place yet in slowing America’s decline, and secure a position in, say, the SEC or DEP.
A career in public service, however, may not appeal to all that many millennials. Given this, carving out a space in the private sector could be the most realistic, if not most rewarding, option. The rise of AI will inevitably destroy a host of jobs (47% of US professions will be eliminated, if the famed Oxford study of 2013 is to be believed). But, with compliments to Schumpeter, a host of new industries should emerge — and on whose coattails millennials would be wise to ride. While many of the jobs so created will be quite technical in nature (and thus suited only to the minority born with the right neurotype) I suspect that the majority of these new jobs will require only passing technical fluency.
My intent in the following is to sketch out a few concepts that fall into this latter category, i.e. nascent industries ripe for entrepreneurial intervention and poised to benefit from the rise of automation. These concepts are admittedly on the idealistic side, tending as they do toward the employee-centric and externality-averse. Nevertheless, I believe them to be viable business ideas with reasonable chances of success in the American marketplace.
A. Artisanal production
This is a lifestyle (non-scaling) company direction, marking a return to the master craftsman. In this scenario, one would achieve mastery of a particular trade, whether through apprenticeship (similar to that practiced in medieval guilds) or self-teaching. Goods produced, whether individually or collectively, might include furniture, true luxury homes, appliances, shoes, or urban gardens. To maximize their appeal to consumers, the goods would be bespoke and of supremely high quality.
While a master craftsman’s toolkit would include traditional tools, they might also leverage more recent production techniques such as 3D-printing and laser cutting. And though there are many drawbacks to the artisanal production approach (e.g. lack of apprenticeship opportunities, years of painstaking training, difficulty of self-marketing once proficiency is attained) I believe that there is substantial unmet demand, especially from the middle class and up, for authentic products whose consumption benefits individuals rather than multinationals. The success of companies like Etsy and educational institutions like ACBA bear out this belief.
B. Mass production
This company direction results from extending the former concept to mass markets. (While Etsy is successful at what it does, it is unable to supply 99.99% of the goods required by ordinary consumers.) Indeed, rather than selling one-offs, in this scenario one would build a company that specializes in mass-production of a specific line of goods — say, kitchen appliances, kids toys, or office furniture. To maximize their appeal, the products would be domestically manufactured and forefront their craftsmanship — raw metals, polished woods, sumptuous housings. The value prop would follow from the quality, story, and ethics (e.g. no sweatshop labor) of the goods. To bring costs down to a reasonable level, one could cut out overhead and intermediaries at every stage of the product development cycle — pursuing distributed manufacturing, distributed assembly, and a distributed workforce. All-in-one 3D-prints offer a promising cost-cutting option as well.
Retailers as diverse as Home Depot, Williams Sonoma, and Sur La Table, interested as they are in offering compelling upmarket options, might be convinced to stock these goods in certain of their locations. In closing, I should admit that this business model is little different from the commercial model of 1950’s America. In fact, its direct antecedent would be William Morris’ Arts and Crafts movement, which arose in the 1850s under similar circumstances.
C. Riding the blockchain
Blockchain apps will enable new modes of consumption and commerce, allowing for micro-transactions (of currency, energy, fungibles, and much more), secure debiting and crediting, and streamlined contract enforcement. While I’m not as optimistic as some regarding the ramifications of blockchain APIs, I believe that many entrepreneurs will benefit enormously from the freedom and power conferred by them. As blockchain smallholders line up along the long tail, one should expect a concomitant rise in related open source tools. For instance, Opendesk serves up blueprints for common pieces of furniture, then (pending payment) connects you with a local woodshop capable of manufacturing them for you. This business model could be extended to any number of goods, from bespoke software to custom orthotics. I suspect that the sweet spot here could consist in the IKEA-ification of larger goods. For instance, rather than requiring manufacturers to fully assemble large objects (which is costly from a labor and shipping point of view), an entrepreneur could flat pack all its ingredients, and ask that consumers carry out final assembly in the comfort of their home. On this accounting, mile n > mile n-1.
The efforts of behemoths like IBM to mainstream blockchain technology ought to help entrepreneurs interested in making a living off of it. Contract creation and enforcement is a perennial bane for small business owners, as startup and contract and arbitration lawyers suck up immense amounts of time/money/energy. Replacing them with automated tools trustworthy to all parties would mark a major advance. Disintermediating expensive but dispensable knowledge brokers will, in the best case, be blockchain’s great contribution. (Though this value-add is certainly still up for debate.) In any case, the pressure to deliver real value of this sort will be doubly important once the Bitcoin bubble bursts and crypto currency speculators rush to the exits. I expect that Ethereum, or a similar platform with broad-based appeal, will before long (thanks to network effects) become the bedrock of this approach to entrepreneurship.
D. Workforce training initiatives
The most needed of the ideas here, workforce training initiatives would connect existing employers with job-ready employees. The chief complaint of many companies today is the lack of qualified candidates to fill entry- and mid-level roles. At the same time, few companies have the resources or expertise to adequately train new hires. The intent of this business model, therefore, is to equip low-skill workers with both the general and specific skills required to excel in the jobs companies need. One obvious focus would be machining and heavy industry, where there is a major shortage of qualified workers.
A challenge to this approach is the difficulty of crafting a custom curriculum that teaches skills in a memorable way. Another is teaching young people that have been failed by the public education system, and consequently lack basic numeracy and literacy. Learning from the successes and failures of MOOCs will be instrumental in bringing to market a solution of this sort. The upside is that there is significant demand for such a service, and at that from companies in a position to pay handsomely for it.
E. Smallholder Farming
If there’s one constant in American supermarkets, it’s that the food for sale was grown on industrial farms with excessive pesticides. Whether one looks at vegetables, processed foods, or meats, the result is the same: an edible with measurable quantities of chemicals never intended for human consumption. One of the only ways to (potentially) escape this is to buy foods from a farmers market. Farmers markets, to those with the requisite time, money, and zip code, are also one of the only places where one can obtain flavorful produce; industrial farming in America involves the use of seeds genetically engineered for resilience, not flavor.
One response to this situation is to follow in the footsteps of Helen and Scott Nearing and begin small-scale farming. This path is neither easy nor lucrative. But it does afford one the ability to lead a self-sufficient life (not to mention a degree of security in the event of a catastrophe). And there are many resources today that were not available to the likes of the Nearings when they began their farming experiments in the 1950’s.
More specifically, there are cheap and fast-deploying greenhouses that enable year-round farming; new agricultural methods, such as hydroponics; new labor-saving devices for planting, monitoring, watering, and harvesting; and better methods for connecting with supermarkets and individual consumers. Cash flow positive agriculture startups like RoBotany, Grove, and Iron Ox validate the vim of this market. New techniques for growing artificial meats are extremely promising as well. Many affluent consumers would readily switch from real to artificial meat, if the only perceptible difference were the price tag.
IV. Conclusion
My research has made two things clear: the most probable futures of the G7 labor market are mixed, but cloudy on the whole; and there are no silver bullets when it comes to solving the problem of a labor surplus. On the first point, it has been surprising to see how long-established cycles — viz., the devolvement of power from states to cities, the ripple effects of globalized trade, popular unrest accompanying technological unemployment — persist and indeed become more salient month by month. Also, the sanguine appraisals of many experts regarding the future of AI-driven technological unemployment seem frighteningly ill-considered. Net unemployment in the US may indeed not dip below 6% — GDP may hold steady at 2% — the Dow may post positive returns year on year — but this prosperity is all beside the point if half the country doesn’t share it. This is not a moral position so much as a pragmatic one: aggrieved voters will tear this Republic to pieces before the winners of the new game have any chance to enjoy their spoils.
On the second point, I believe it will be incredibly hard to create jobs for all the workers who are displaced by AI. Keynesian stimulus and federal labor market intervention is the closest we have to a panacea, but such solutions are unlikely to ever see the light of day. The best hope for the US is to foster entrepreneurship, and in particular to make way for new businesses that leverage the AI-ified economy, while at the same time maximizing employment of the non-technical majority. The ideas proposed here are not by themselves up to the task of taking up the slack. But they may well provide a livelihood — and, if not more important, an existential satisfaction — to the many who will find themselves jobless in the wake of AI. My hope is that not just well-meaning but self-interested millennials will pursue these new directions, and in the process build a labor market that is more equitable and robust.
