The world needs to talk about the Lamplighter Problem
on the broader ramifications of AI and mass job displacement
For we are all very lucky, with a lamp before the door
And Leerie stops to light it as he lights so many more.
- Robert Louis Stevenson, “The Lamplighter” (1885)
I wonder what it was like to have a friendly lamplighter like Leerie in the neighborhood, making the rounds each evening to manually light the street lamps. Stevenson’s nostalgic view of the lamplighter is similar to how older generations may remember milkmen — friendly faces and unofficial watchmen in the neighborhood. By the early 20th century, lamplighters like Leerie lost their jobs as gas lamps were replaced by electric lighting. Some were retrained for electrical work; others moved on to industrial and municipal jobs.
Today, hundreds of millions of people face a similar fate as the lamplighters. McKinsey estimates that AI-driven automation could displace 400 to 800 million global workers by 2030.1 As someone at the intersection of technology and politics — and also a dad — I look toward the future and wonder: What will happen to humanity as smarter machines take over? How can society cushion the impact?
We are all “Leeries.” AI is ushering in a labor revolution as profound as the Industrial Revolution, affecting every role in the global economy — port workers, lawyers, designers, and more. The stakes couldn’t be higher: the fate of the economy, democracy, and humanity itself. The problem is that public discourse is lagging far behind this revolution. It’s like a tsunami is approaching while we sip Mai Tais, discussing last week’s weather. If we don’t catch up fast, AI will reshape the world before we even realize it, and the revolution will be a fait accompli before we can respond. By raising awareness and fostering discourse now, we still have the opportunity to shape the AI revolution and mitigate its effects, ensuring diverse perspectives are incorporated beyond the few driving the technology. We must be proactive not reactive. It is imperative that we humans stay in control of our destiny.
As part of this, we need a simple way to talk about the sweeping challenges ahead — a shorthand. That’s why I am branding it the Lamplighter Problem. Inspired by Stevenson’s poem, it captures the massive, tsunami-like displacement coming our way. The lamplighter metaphor works because it is removed from present-day concerns, symbolizes someone who literally brightened the world through their work, and humanizes both sides of the displacement.
To define the concept more precisely, the Lamplighter Problem refers to the cascading challenges posed by automation, where AI not only reshapes the economy but also disrupts politics, social structures, and the human experience. It emphasizes the broader impact of mass job displacement, focusing on ramifications that extend far beyond economics.
Both Sam Altman and Curtis Yarvin have recently written about these issues through the figure of the lamplighter. Altman, CEO of OpenAI, referenced the lamplighter in his essay, “The Intelligence Age,” which paints an optimistic view of AI’s potential for humanity.2 “Nobody is looking back at the past, wishing they were a lamplighter,” he noted.
In contrast, Yarvin argues in a response titled “Sam Altman’s Lamplighter” that we should wish more people were lamplighters today.3 To Yarvin, it is better to be a lamplighter than an idle welfare recipient collecting UBI checks and binge-watching Netflix. Work, in his view, gives people meaning, dignity, and purpose. He advocates creating traditional, artisanal jobs to counter automation’s destabilizing effects, a sort of large-scale Etsy economy. I’m not certain that’s the solution, but I strongly agree with his core point: automation isn’t just an economic problem — it’s a political one. Yarvin rightly points out that displacement caused by the Industrial Revolution gave rise to communism. If we push automation too hard without addressing its impacts, the political consequences could be dire. Enslavement by robots? Fully automated Stalinism? No thanks!
As I read their essays, I couldn’t help but picture Leerie and the many “Leeries” in my life today. I know both Altman and Yarvin, and I’m friendly with many in the tech world. At the same time, I have loved ones whose jobs could someday be automated, while others are question marks. Will my nephew beginning his career as a commercial pilot be displaced by autonomous planes someday? I just don’t know. I feel anxious thinking about it. So I find myself of two minds: AI excites me, and I want to accelerate into the future. But I’m also wary of the upheaval it will cause. This transformation could displace millions, destabilize lives, and challenge the structures that give people meaning and purpose, including democracy itself. This personal conflict mirrors the broader societal tension, which is why we need a shorthand like the Lamplighter Problem to discuss these challenges.
At its core, the Lamplighter Problem asks us to consider how to maximize human flourishing and empowerment in the coming “intelligence age.” How do we make sure the AI tsunami benefits humans? How do we make sure these benefits are shared? That’s my north star, and I suspect many would share it, though it’s worth debating. But even with a shared north star, people will have different ideas for achieving it. For some, advancing human flourishing will mean standing in the way of AI advancements yelling STOP. For others, it will mean accelerating AI at full throttle, yelling GO GO GO. Most of us will find ourselves somewhere in between. My approach leans toward GO, but with pragmatic caution.
This isn’t theoretical. We already see versions of the Lamplighter Problem in action. Take, for instance, the recent labor dispute involving Harold Daggett of the International Longshoremen Association, a key figure in the fight over port automation. At first, I thought: Automate the ports! To hell with this guy and his union for holding America’s supply chain hostage before an election. Economically, automating our ports as they did decades ago in Rotterdam is a no-brainer. But when I consider the thousands of jobs at risk, the ripple effects across families and communities, and multiply that across hundreds of other professions, the Lamplighter Problem becomes clear. It reminds me that these issues require a humble, thoughtful approach.
As a society, we must face the Lamplighter Problem head-on. We have to talk about it. By naming it, I hope we can foster a conversation that looks beyond the horizon to the social and political fabric of human flourishing. But that’s just one step. We need to broaden awareness and advance the discourse on these issues, from barstools to legislative halls. I will do my part. In a future essay, I will map out different schools of thoughts and various solutions that have been discussed so far.
In the meantime, maybe it’s wise to let Leerie light the lamps a few more nights.
Special thanks to and others for feedback on drafts. I want to know your thoughts! Please share in the comments. Below is the full poem.
The Lamplighter
by Robert Louis Stevenson (published in 1885)
My tea is nearly ready and the sun has left the sky;
It’s time to take the window to see Leerie going by;
For every night at teatime and before you take your seat,
With lantern and with ladder he comes posting up the street.
Now Tom would be a driver and Maria go to sea,
And my papa’s a banker and as rich as he can be;
But I, when I am stronger and can choose what I’m to do,
Oh Leerie, I’ll go round at night and light the lamps with you!
For we are very lucky, with a lamp before the door,
And Leerie stops to light it as he lights so many more;
And O! before you hurry by with ladder and with light,
O Leerie, see a little child and nod to him tonight!
I highly recommend Noah Smith's piece on why comparative advantage means AI will most likely be reserved for its most efficient uses. Same reason why Warren Buffett has a secretary: It's more valuable for him to spend his time making money.
https://www.noahpinion.blog/p/plentiful-high-paying-jobs-in-the
AI as productivity software will probably streamline a lot of "email jobs", and to a large extent that's a valuable cost savings, but if the next generation of AIs unlocks an insane improvement in protein folding that revolutionizes biotech, most of the compute will be spent on that.
I've been nursing a thesis lately that even with widespread use of AI as productivity software, humans will just march on up Maslow's pyramid and "fake it more than ever" rather than actually instituting UBI:
(1) WFH will become more ubiquitous and come to resemble something like a true "8 hours of real work / 32 hours of virtually pretending we're working = 40 hours to legally satisfy outdated labor-era regulations" as AIs optimize our schedules while we go to the grocery store.
(2) With AI tutors, more students qualify for college education, and college-level instruction becomes cheaper and more unmoored from expensive campus experiences than ever. Instead of a permanent UBI-dependent underclass, we get a larger PMC than ever, nursing those email jobs on "8/32=40" schedules. It's a leisure society the way Keynes always imagined it, except for those 32 hours we're still doing light knowledge work and organizing around those 8 hours.
(3) Because of all this, no one ends up *needing* a UBI. Job retraining becomes cheap and easy with AI tutors specialized for adult workers changing careers. Job allocation becomes easier with AIs tasked to sift through thousands of job listings for us.
The "UBI" really just ends up being people paid to sit and think up our next level of needs all day long.
"To Yarvin, it is better to be a lamplighter than an idle welfare recipient collecting UBI checks and binge-watching Netflix."
I agree, but if we respond to a fall in real wages in many occupations with more generous EITC rather than UBI we avoid this dilemma.