The remarkable success of ChatGPT and other generative A.I. has inflamed the already roiling debate about how the rise of machines will affect workers, and ultimately shape our societies. Doomers predict that robots will replace people and, perhaps, even destroy the human civilization. Optimists argue that, after the inevitable growth pangs, new intelligent machines will make our life better. After all, humanity has successfully digested previous technological revolutions without particularly dire consequences.
But learning from history is not a straightforward matter. The A.I. revolution will impose new and unanticipated stresses on our society. A key dimension missing in the current debate is the role of social power, which determines not only the winners and losers of such technological shifts, but also how much social and political disruption will come in the wake of a technological shift.
An extreme illustration of this principle is the fate of one special class of workers, the horse. As the Nobel Prize-winning economist Wassily Leontief pointed out forty years ago, between 1900 and 1960 the equine population in the U.S. collapsed from 21 million to only 3 million.[i] The culprit was the internal combustion engine, which enabled the automobiles and tractors to replace horses in transportation and agriculture.
The horse is a physically powerful animal, but its social power is zero. Horses’ life and death are entirely under control of humans. As a result, the technological shift brought about by the internal combustion engine resulted in, essentially, horse genocide. But at the societal level, the demise of the workhorse caused hardly a ripple—no horses rebelled on the way to the glue factory.
What about humans? Consider the “workhorse” of the industrial era—men without college degrees. They fared better than horses, but worse than other segments of the population. The proportion of American men aged 25-54 in the labor force has declined by nearly 10 percent from its peak in the late 1960s.[ii]
Real wages of men without college degree have stagnated since the late 1970s. More recently, life expectancy of men without college degrees has decreased.[iii]
Of course, technological change was only partially responsible for this disturbing development. I am making a larger point here—about the key role of social power. Recent research by economists provides strong evidence that declining worker power has been a more important factor in depressing their wages than the interplay of labor supply and demand.[iv] The erosion of collective bargaining rights, weaker labor standards, and new employer-imposed contract terms together account for the bulk of the divergence between productivity and median hourly compensation growth.[v] There is a growing consensus among economists that unequal power has played the most important role in suppressing pay increases for non-elite workers since the 1970s.[vi]
An important dimension in the declining social power of American workers is the evolution of the Democratic Party, which was a party of the worker class during the New Deal, but by 2000 became the party of the credentialed “ten percent”.[vii] The rival party, Republicans, primarily served the wealthy 1%. The 90% were left out in the cold. Amory Gethin, Clara Martínez-Toledano, and Thomas Piketty studied hundreds of elections and found that political parties in other western democracies also increasingly cater to only the well educated and the rich.[viii]
Workers didn’t accept their immiseration meekly. There are many signs of growing popular discontent. But history teaches us that revolutions are not made by the “masses.” Popular immiseration and resulting discontent needs to be channeled against the governing regime, and this requires organization by breakaway elites, so-called “counter-elites”.[ix] Without such leaders and organization, non-elite workers have been reduced to voting for populist, anti-systemic politicians, like Donald Trump.
So much for history. What about the future? The rise of intelligent machines will undermine social stability in a far greater way, than previous technological shifts, because now A.I. threatens elite workers—those with advanced degrees. But highly educated people tend to acquire skills and social connections that enable them to organize effectively and challenge the existing power structures. Overproduction of youth with advanced degrees has been the main force driving revolutions from the Springtime of Nations in 1848 to the Arab Spring of 2011.[x]
The A.I. revolution will affect many professions requiring a college degree, or higher. But the most dangerous threat to social stability in the U.S. are recent graduates with law degrees. In fact, a disproportionate share of revolutionary leaders worldwide were lawyers—Robespierre, Lenin, Castro—as well as Lincoln and Gandhi. In America, if you are not one of super-wealthy, the surest route to political office is the law degree. Yet America already overproduces lawyers. In 1970 there were 1.5 lawyers per 1000 population; by 2010 this number had increased to 4.[xi]
Too many lawyers competing for too few jobs didn’t depress all lawyer salaries uniformly. Instead, competition created two distinct classes: winners and losers. The distribution of starting salaries obtained by law school graduates has two peaks: the right one around $190k with about a quarter of reported salaries, and the left one around $60k with about half salaries, and almost no salaries in between.[xii] This “bimodal” distribution evolved from the usual (“unimodal”) one in just one decade, between 1990 and 2000. Here’s what it looked like by 2010:
What this development means is that those in the right peak have succeeded in getting into the pipeline to elite status. Most of those in the left bulge, on the other hand, will become failed elite aspirants, especially considering that many of them are crushed by $160k or more of debt they incurred to pay for law school.[xiii]
If the outlook for most people holding new law degrees looks dire today, the development of new A.I. will make it much worse. A recent Goldman Sachs report[xiv] estimates that 44% of legal work can be automated—lawyers will be the second worst-hit profession, after Office and Administrative Support. If market forces are allowed to have their way, we will create a perfect breeding ground for radical and revolutionary groups, feeding off the vast army of intelligent, ambitious, skilled young people with no employment prospects, who have nothing to lose but their crushing student loans. Many societies in the past got into this predicament. The usual outcome is a revolution or a civil war, or both.[xv]
Unless the worst fears of the doomers are realized, in the long run, undoubtedly, we will learn how to race with smart machines, rather than against them.[xvi] But in the short-medium term (say, a decade), the generative A.I. will deliver a huge destabilizing shock to our social systems. Since we can foresee the effect of ChatGPT and its ilk on potentially creating large numbers of counter-elites, in principle, we can figure out how to manage it. The problem is that I have no confidence that our current polarized, gridlocked political system is capable of adopting the policy measures needed to defuse the tensions brought about by elite overproduction and popular immiseration. I hope I am wrong; but if not, prepare for a rough ride.
[i] Brynjolfsson, E. and A. McAfee (2015). Will Humans Go the Way of Horses? Foreign Affairs. New York, Council on Foreign Relations NY. 94: 8-14.
[ii] https://fred.stlouisfed.org/series/LREM25MAUSA156S
[iii] Case, A. and A. Deaton (2020). Deaths of Despair and the Future of Capitalism, Princeton University Press.
[iv] Stansbury, Anna and Lawrence Summers. “Declining Worker Power and American Economic Performance,” Brookings Papers on Economic Activity, March 19, 2020, https://www.brookings.edu/wp-content/uploads/2020/03/stansbury-summers-conference-draft.pdf.
[v] Mishel, Lawrence and Josh Bivens, “Identifying the policy levers generating wage suppression and wage inequality,” Economic Policy Institute, May 13, 2021, https://www.epi.org/unequalpower/publications/wage-suppression-inequality/.
[vi] Noam Scheiber, “Middle-Class Pay Lost Pace. Is Washington to Blame?” New York Times, May 13, 2021, https://www.nytimes.com/2021/05/13/business/economy/middle-class-pay.html.
[vii] Frank, T. (2016). Listen, Liberal: Or, What Ever Happened to the Party of the People? New York, Picador.
[viii] Amory Gethin, Clara Martínez-Toledano and Thomas Piketty, “How politics became a contest dominated by two kinds of elite,” The Guardian, August 5, 2021, https://www.theguardian.com/commentisfree/2021/aug/05/around-the-world-the-disadvantaged-have-been-left-behind-by-politicians-of-all-hues.
[ix] Turchin, P. (2016). Ages of Discord: A Structural-Demographic Analysis of American History. Chaplin, CT, Beresta Books.
[x] Goldstone, J. A. (1991). Revolution and Rebellion in the Early Modern World. Berkeley, CA, University of California Press.
[xi] Ages of Discord, Figure 4.5.
[xii] Data for 2020 from National Association for Law Placement, NALP.
[xiii] https://www.americanbar.org/content/dam/aba/administrative/young_lawyers/2020-student-loan-survey.pdf
[xiv] https://www.ansa.it/documents/1680080409454_ert.pdf
[xv] https://seshatdatabank.info/
[xvi] Brynjolfsson, E. and A. McAfee (2016). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. New York, WW Norton.
After reading your brilliant End Times book, and your thought provoking look at AI, it’s obvious that the next 10 years are going to be even more horrific than what happened during World War 2. The question is, how can we best prepare and fortify ourselves and our loved ones for what is coming?
“The problem is that I have no confidence that our current polarized, gridlocked political system is capable of adopting the policy measures needed to defuse the tensions brought about by elite overproduction and popular immiseration.”
Even within an ideal system, how could these tensions be defused? At most, elite overproduction could be stopped, but then there’d still a lot of surplus elite (which will take at least decade to vanish by retirement of present elite). What could be done with that surplus elite to keep them from causing problems?
Create colonies and give them elite jobs in colonies? I don’t see any feasible solution.
Downshifting/leaving the rat race, state-created jobs, turning off the wealth pump (see explanation in End Times), and, yes, colonization of the Solar System (why not?)
Thanks Peter for such an inspiring start to Thanksgiving week (that’s snark for the humor impaired). Seriously, I think we’re in for a hell of ride politically, culturally, and socially not just in the U.S. but worldwide because of the dynamics you’ve illuminated here. I just hope the world doesn’t blow up in the process.
Look at where the research money is going: math, physics, business, medicine, economics, psychology (see the recent State of AI report). Combine this with well-paid professions. It’s really easy to see the targets: lawyers, accountants, management consultants, i-bankers, therapists, even most physicians. Careers not to pursue. Follow the data …
I agree with you. Just from the politic news from the US it seems clear that something has to change and AI might be the trigger.
There are two things that from a structural-demographic point of view argue against serious instability in US (or UK). First is the age structure, there is not a large enough cohort of inflamed youth. Second there is no revolutionary narrative. We have concern for the marginalised, and fear of climate change. But no narrative of a brighter future: in the manner of Animal Farm: “Beasts of England, beasts of Ireland/Beasts of every land and clime/Hearken to my joyful tiding/Of the golden future time” . Of course, as Turchin noted, the unfortunate loyal; horse Boxer went to the knacker’s yard.
I agree about age structure. The most recent youth bulge peaked during the late 2010s, and currently the number of people in the most “dangerous” age cohort is declining. But the youth bulge is only one, and not the most important driver of instability.
Thank you for replying. In the 25th anniversary edition of his Revolutions book Jack Goldstone says he understated the role of ideology:
Difficult question.
Can a true populist – one pushing for reforms like taxing the rich – get into power in the US by lawful means at this point? It seems like the elite class – of both parties – erects tremendous barriers to entry. I want peaceful reform, but it seems unlikely.
It happened before — the Progressive/New Deal eras. Why not today?
Professor Turchin’s comments on “counter-elites” evoke what philosopher Eric Hoffer wrote in his acclaimed 1951 volume “The True Believer,” about “Men of Words.” The first sentence in Chapter XV “Men of Words” states: “Mass movements do not usually rise until the prevailing order has been discredited. The discrediting is not an automatic result of the blunders and abuses of those in power, but the deliberate work of men of words with a grievance.” (Obviously today we would replace “men” with “people.”)
Mass movements + dissident elites is what achieves change
I have a somewhat insider view on what happened since the 1970’s because I provided engineering and construction services for literally all the major industrial firms ranging from GE to Kellogg in every sector from pharmaceuticals to power generation in the US and in China, Australia, Canada and South America. All of these corps moved their operations out of the US for two reasons, both of which significantly impacted their bottom line:
1) Cost of environmental compliance due to ever increasing stringent regulations
2) Cost of labor due to labor unions.
Thanks Peter! I see a lot of applications for AI in other fields but I agree that the most socially dangerous bunch will be underemployed lawyers.
As an observation, a lot of the work that I did as a geologist could be done more efficiently by an AI bot, although field work might be a challenge for it. I hope to live long enough to see the technology mature
And the way I work today is much more computer-reliant compared to 40 years ago, when I was in grad school. Much of it has happened very gradually. E.g., word-processing — spell-checking — ChatGPT was a continuous evolution, rather than a sudden switch to AI
Interesting case for (or against) AI. It remains to be seen whether the emergence of AI as a major economic force will have qualitatively different effects on society from the emergence of the world wide web in the mid-1990s or the emergence of the microcomputer (pc) before that. For example, perhaps increasing efficiency in the law field will manifest mostly as reduced employment of lower echelon roles in the profession (paralegals, various support staff, etc.). One question is the graph of the bimodal income distribution for law graduates played up in the posting above (and in End Times, and in Ages of Discord). The graph is old. What has the salary distribution been looking like in the decade plus since then? Same problem as with End Times overall – little updating of older data already presented in Ages of Discord.
The most recent data I’ve seen suggests that the peak on the right (high earners) keep moving to the right, but is also getting smaller. It’s still bimodal.
Perhaps one hoped for escape valve with respect to lawyer and elite over-production is if lawyers ever figure out how to make money off class action lawsuits from civil disturbances (like road closures). That could function as a societal stabilizer. Right now political disfunction & lawfare seem to be the new expansion market – and that marketand its dynamics are obvious destabilizers.
That would be even more a dystopia than what happens in the US already… because it would intensify lawfare, rather than diffuse it, wouldn’t it?
very oustanding viewpoints!
Regarding revolutionary lawyers let us not forget to add Nelson Mandela to the list.
Yes
As revolutions, violent or democratic, need to be financed it will be some “conservative”, ie. market liberal revolution, further exacerbating the inequality problem. On top of that, there is the transition of capitalism to finance capitalism (bigger turnover is made by dealing in finance than in whatever other goods and services) which by itself is always an endpoint (van Bavel, “Invisible Hand?”, s. a. review by Branko Milanovic). Then there are increasingly global demographic problems, and we don’t want to think about potential climate issues. Potentially there may be a dark century or two ahead. I would not count out China, where the government still enjoys lots of goodwill and therefore has some room for maneuvering.
This article reminds me of the songs “Democracy” and “Waiting for a miracle” both by the late poet song writer Leonard Cohen
The elite are only relevant because men with guns obey. Once robots with guns are on the scene, a leader no longer needs an elite to enforce their will. Once the machines no longer require humans to maintain and supply them, internal human resistance ceases to matter.
There is no stopping the social dynamics; the elite dumped virtue so they are unable to deal with the promise of ultimate power in a cooperative manner.
[…] When A.I. Comes for the Elites by Peter Thurchin. […]
So much legal work is being offshored. When I get asked to approve terms on a contract, it comes back with comments from India. That would have been done by a young American lawyer in the past. I expect a lot of recent and future law grads will end up going into disputes, which would always involve negotiation with human parties or appearing in a courtroom in front of a judge. It will be hard for AI to replace that. Meanwhile, LLMs will be a freaking godsend for discovery and will likely generate dozens more theories/causes of action on any complex case where they are used. This might end up being a boon for lawyers.
The bigger impact from AI will be the complete loss of trust; capping a 4 decade process that began with the “free and open” permissionless Internet of the early 1990s. The lack of any type of settlement system of economic incentives and disincentives completely imbalanced risk; almost entirely toward the receiver. And that has resulted in the societal, technical and financial ills plaguing us today. And on top of that, AI is simply a natural and much more alarming progression of the big grift (content, capital, power, etc…) that began with that free and open internet; which was anything but in hindsight. By 2030 we will look back and realize that the loss of trust outweighed discontent among the elites; if we make it that far.
In the US, military service as an officer is entree into a higher class (consider the phrase “an officer and a gentleman”) and both it and service in the upper echelons of the security services like the CIA are also strong credentials for a political career. The online World Socialist Website likes to keep track of the “CIA Democrats” as it likes to call them.
In addition to being an officer, family connections in politics are also a strong credential for political office. Consider the state of WV in the US. Of the two senators, Joe Manchin is the nephew of a long-time political operative, Secretary of State A. James Manchin. The other, Shelley Moore is the daughter of one-time governor Arch Moore. (Married name Capito, but the Moore is the one that counts.) One of the representatives, a Carol Moore is the daughter of an Ohio congressman as well as the wife of a major car dealership owner. That of course signals the other great road to office, personal or family wealth. The current governor Jim Justice is held to be a billionaire. A previous governor, Gaston Caperton, was what counted in Charleston as old money. The aforementioned Arch, Shelley’s daddy, is said to have kept thousands of dollars in cash in his desk drawer, but in WV politics, the saying is, indictments don’t count. In most rural areas, the largest business owner and/or landlord are almost always highly significant figures. When it comes time for the local lawyer who held elective office to invoke the community/the people, it is remarkable how often their respectable citizens are the pharmacy owners or a bank president or a physician who owns a medical building or a mine operator.
It is not always very clear in what sense merely possessing a higher education degree, even in law, makes one an elite. And the tag counte relite really does smack of, traitor to their class. But even for a law degree, with the already prominent bimodal distribution of income, I’m not sure they are in the elite enough to be counter, as opposed to mere failed aspirants. Historically, one of the greatest rebel leaders, although ultimately unsuccessful in conquering the whole of China, was Hong Xiuquan. Another great revolutionary leader (and yes, La Reforma *and* the revolution against Emperor Maximilian counts!) was Zapotec in nineteenth century Mexico. That is not elite, even if he trained as a lawyer. Wealthy men and military officers who acquired a law degree but never really practiced are part of the elite but attributing that status to their degrees rather than their money or the status as officers? Even more, how many of these people were using their connections, their elite status, rather than legal expertise? As I understand it, when Edwin Stanton worked with Abraham Lincoln, Stanton was the big shot who had no use for the mere lawyer who only had legal knowledge and rhetorical skills. Indeed despite all those law degrees, who really worked as a lawyer besides Lincoln? Wilson was highly credentialed true but something of an electoral fluke (Taft vs. Teddy Roosevelt split the Republicans.) Otherwise, aren’t most national leaders more family and money and family money than credentialed?
What is missing from this equation, is the climate crisis. If the majority of climate experts are correct, mass explosions of all types will erupt once the climate passes the tipping point. Then A.I. may well be meaningless. Even if either the elites or counter-elites are able to assume some type of control over it, their efforts will be directed only toward a small minority, not humanity overall.
Oh, lots of things are missing. The risk of nuclear holocaust has grown over the past several years.
While I must express my disagreement with a degree of deference, it is my considered view that AI stands as a potential cornerstone in addressing the multifaceted challenges posed by climate change.
To elucidate, I refer you to a scholarly article from the June edition of the American Chemical Society’s Environmental Science & Technology journal. This piece, entitled “Machine Learning in Environmental Research: Common Pitfalls and Best Practices,” is a nexus for a considerable corpus of nearly 50,000 related academic papers. While the sheer volume of these papers is noteworthy, I would direct your attention more specifically to the lead author’s page at Princeton University. Here, one can engage with the fingerprint of this paper. Additionally, for a more interactive exploration, see the relevant Litmaps page.
In summation, the role of AI in unraveling and tackling our environmental and ecological quandaries cannot be understated. This is a prime illustration of the benevolent potential of AI. AI for good.
Let’s develop combined human-AI systems — superminds — to tackle our ecological challenges. See this PDF and this HTML version of the PDF, a paper co-authored by several researchers at MIT’s Center for Collective Intelligence.
Thank you for this post, in particular the points of reference it provides access to.
I do not disagree with what you have written, as humans have in the past shown a great capacity for using various tools to help solve their various conundrums. While doing so, however, they oftentimes create new, unanticipated, problems. A.I. may well help in that regard, as the computing capacity of machines to address a multiplicity of concerns is quite noteworthy, and thus that ability may not only bring attention to, but also present various scenarios, of dealing with, such concerns.
My main issue relative to the climate crisis is the human aspect. We humans are not well known for our ability to think long-term–well, perhaps I should have stated we are not well-known for our ability to act in a “positive” manner relative to any long-term thinking. I would argue that, when that is coupled with humanity’s proclivity toward our tendency to act in a parochial fashion relative to which of our fellow humans should receive our support–in any endeavor–history has shown that the two combined do not bode well in terms of results.
In essence, then, I am arguing that the above two-listed items may well prevent humanity from being able to achieve any meaningful action relative to the climate crisis, before the climate crisis overwhelms our ability to do so.