Robots, Razors, Cameras, and Screwtape
Off-Ramps for 29 May 2025
Welcome to Off-Ramps! Today I’ll highlight four interesting pieces that I think you will enjoy reading. Please enjoy all of these on your morning commute, or save them for your weekend.
Robots on the Waterfront
(like me, a 2024 RPI Fellow) delivers a sharp indictment of America’s ports. In Jordan’s view, what once was the primary function of the ports—to serve as a node in the shipment of goods to and from the USA—is now only their secondary function. Their first purpose is now to enrich a rent-seeking cartel: the International Longshoremen's Association (ILA), which uses its status at the middleman between shippers and ships to extract extraordinary wages. That alone would be bad, but what’s worse is that the ILA explicitly blocks the introduction of automation at its ports. More technology would reduce the ILA’s leverage, so that technology remains unused.McGillis’ analysis is strong on the political tensions at play. The coalition that brought President Trump to power included techno-optimists who believe in the power of technology to bring about abundance, and as such would favour making the USA’s ports the best in the world (or, put another away, Great Again). But that coalition also includes the ILA, and so far Trump has backed the latter hard: “I’ve studied automation, and know just about everything there is to know about it,” he posted on Truth Social. “The amount of money saved is nowhere near the distress, hurt, and harm it causes for American workers, in this case, our Longshoremen.”
Like Jordan, I deplore the ILA’s eagerness to keep the USA stuck in the 1970s because that will allow the longshoremen to more easily pick everyone’s pockets. But, as per our philosophy here at Changing Lanes, I think it will take more than technology to bring the USA’s ports into the twenty-first century. Our favourite writer on the subject, Brian Potter, wrote an extensive essay on the subject last year, where he noted that the U.S. actually operates more automated terminals than most countries, yet American ports still perform dismally: Los Angeles, with two automated terminals, ranks #375 worldwide. Nor is automation all that is required. Rotterdam, the poster child for port automation, barely outperforms the much-less-automated New York.
So Jordan is right that blocking inevitable innovation is economically destructive. But Potter is also right that American port dysfunction runs deeper than union obstruction. Other culprits include the facts that U.S. ports don't operate 24/7; don’t coordinate truck arrivals, leading to massive congestion; don’t link rail systems between individual terminals within the same port; and haven’t optimized customs processes.
Jordan’s political analysis is relevant here: the ILA has successfully framed the terms of debate around ‘automation’, positioning itself as defending workers against technological displacement, which means observers line up to argue in automation’s favour. This squabble means that other, more significant and more tractable problems—poor management, inadequate investment, regulatory dysfunction—go unaddressed… even though these problems don’t need to wait on automation. They could be solved without it.
Jordan suggests that the Sunbelt strategy—building up east-coast ports in the south and southeast—might break ILA control. I think he’s right: those ports could not only introduce automation, but also the superior operations that the USA freight network requires. I hope that strategy succeeds.
Roy’s Razor, Cutting Through Hype
Can I sleep in it?
That's it. That's the test. Pick a vehicle. Can you get in, pick a destination and safely go to sleep?
If yes, it's self-driving.
If no, it's not.
I am delighted by
's simple test for whether a vehicle is ‘self-driving’. Like its cousin Occam’s Razor, Roy’s Razor is a useful heuristic for cutting through confusion. Forget the SAE J3016 levels (and if you don’t know what those are, don’t ask); in almost every case, we don’t need technical or regulatory classification. We only need to know what matters most to users: can you truly disengage?As Roy tells it, his mother just wanted to know if she could nap in a Tesla. Her question was natural, given how much hype that Tesla emitted in the past 15 years or so about “Full Self-Driving”. Other companies, like BMW, Mercedes, and Volvo offer similar advanced driver-assistance systems, but all of them fail Roy's test spectacularly. Only robotaxi services like Zoox or Waymo—where Roy notes he “falls asleep all the time”—qualify under this standard.
The language we use shapes expectations, investment decisions, and regulatory frameworks. When everything from adaptive cruise control to full autonomy gets labeled “self-driving”, we create dangerous confusion. Investors and regulators that misstep this way deserve what they get, but when drivers overestimate their vehicle's capabilities, that can cost third parties dearly. That’s why my co-authors and I made the first chapter of The End of Driving (forthcoming in August; pre-order now!) on the subject of language for driving automation: confusion and deception in this space has been rampant, and requires strict policing.
The Inevitability of Public Surveillance
Another RPI 2024 Fellow,
, recently reviewed how the New Orleans Police Department kinda-sorta operated a secret facial-recognition system to identify criminals in public places. In my view, this story is only the latest illustration of how fundamentally incoherent our view of this technology is.What happened in New Orleans is that the city council had established clear rules prohibiting real-time facial recognition, so police found a workaround: an arm’s-length nonprofit took in footage from private cameras, ran the data through a facial-recognition screen, ran results against criminal watchlists, and sent alerts to officers via a private mobile app. In this way, the letter of the law was kept even as the spirit was violated: property owners voluntarily shared camera feeds, an ‘independent’ ‘nonprofit’ ran the system, the NOPD benefited from its alerts, and it all happened without public oversight.
That, to me, is the striking thing about this case: it illustrates the futility of blanket technology bans. The city council's prohibition didn't eliminate surveillance, it only drove it underground, creating a system with even less oversight and accountability than open deployment would have provided. At the same time, it was also less effective, since it could only use some public cameras, those of private operators who consented to participate. To quote myself, from this article from last year: “we have, simultaneously, too much surveillance for a society committed to privacy and liberty, and too little for a society committed to public safety and order.”
The Case for Public Surveillance
To understand how confused our intuitions are about facial-recognition technology are, start with popular culture; start with Batman.
My stance is unchanged: rather than fighting the inevitable expansion of urban surveillance capabilities, which will only improve in an age of artificial intelligence, cities should focus on governance frameworks that ensure democratic accountability. Who controls watchlists? How are false positives measured? What oversight prevents mission creep? These operational questions are the ones that matter, because the question of ‘should there be public surveillance’ is a settled one. As Jeff points out, there will be such surveillance. Standing athwart history shouting ‘stop’ won’t change that fact.
So let’s commit ourselves to using these systems well.
The Screwtape Economy
As a youth I was deeply impressed by C.S. Lewis's The Screwtape Letters, but I haven’t revisited it in decades. So I was delighted to read
’s use of the book as a lens to view our present age.For those who haven’t read the book, it consists of a series of letters written by Screwtape, a senior demon in the bureaucracy of Hell, written to instruct his nephew in the fine art of damnation. If I recall correctly, the central point of the book is that demons don’t need to tempt humans to fall; humans are very good at destroying themselves. It is rather ‘the Enemy’, i.e., the Divine, who is constantly seeking to ‘tempt’ people to look outside and above their own interests and desires. The job of a demon is to distract humans from these opportunities, permitting humans to damn themselves (or, for those who don’t share Lewis’ theological commitments, to destroy not only their potential for goodness, but also the goodness they have already).
Scanlon makes the case that we are living in Screwtape’s economy, a system optimized to do Hell’s work for it. It is a system that, irrespective of the intentions of those who participate in it, systematically hollows out human agency through systematic rejection, frictionless convenience, and algorithmic predictability.
On rejection, Scanlon observes that Harvard accepts only 1,950 of 54,000 applicants anually, while Goldman Sachs gets 315,000 applications for 2,700 internships. Most job postings receive 244 applications on average. In other words, for people of a certain class at least, application and rejection is default condition of economic life. (Scanlon doesn’t mention dating apps in this context, but could easily have done so.) The psychological toll of always being judged, and almost always being found wanting ripples through the rest of life: less empathy, more aggression, and financial nihilism.
Scanlon mentions dating apps in the context of frictionless convenience, which erodes our capacity for difficulty. Screwtape chortles that humans fail to see how small choices, repeated daily, shape our character more than dramatic decisions. Our intolerance for difficulty, expressed in DoorDash’s triumph over going out to eat, is a synecdoche of our age’s impatience; the frictionless economy promises to free us for deeper pursuits but actually atrophies the patience and resilience those pursuits require.
I’ll leave you to explore Scanlon’s views on algorithms for yourself, because it is frictionless convenience that is most relevant to the readers of Changing Lanes. As I have observed on several occasions, driving automation is poised to remove driving’s most obvious cost, namely our attention. And the consequence of that is we drive much more than we do now, while failing to observe that driving’s subtle costs will persist: to our personal health, our engagement with our neighbourhoods, and our relationships with others (not to mention to our possessions).
Screwtape observes that the road to Hell is a gradual slope, “soft underfoot, without sudden turnings, without milestones”. That description applies just as well to one path we might take: into automated-vehicle-dependent, convenience-optimized environments that would diminish our communities and ourselves. Turning off that path will require wisdom.