109 | Jason Torchinsky on Our Self-Driving Future
What this covers
Jason Torchinsky, a senior editor at Jalopnik and author of Robot, Take the Wheel, joins Sean Carroll to examine why the future of autonomous vehicles is less a technical problem than a human one. Rather than walking through engineering milestones, the conversation centers on three intertwined claims: that level two autonomy—the most advanced system commercially available today—is fundamentally incompatible with human psychology; that fully autonomous vehicles will cease to be "cars" in any meaningful sense and become instead reconfigurable mobile rooms; and that society stands to lose something irreplaceable in the shift away from human driving.
The discussion moves across concrete technical ground before pivoting to cultural and philosophical terrain. Torchinsky explores how AI as mimicry of intelligence fails in urban driving, where humans rely on folk rules of driving—eye contact negotiations, gesture conventions—that no traffic code captures. He traces how system-one versus system-two cognition explains why demanding constant readiness in level two autonomy conflicts with human nature: when a machine handles 80% of the work, people disengage rather than remain alert. The conversation addresses whether Tesla's labeling of semi-autonomous systems misleads consumers, examines the pareto principle at work—the final 20-30% of capability proving hardest—and considers how private ownership will persist in car-dependent regions while shared fleets dominate dense cities.
Torchinsky argues that the deepest loss is not technological but existential: driving is one of the last refuges from productivity culture, a space where the body and hands stay occupied while the mind wanders free. An autonomous commute, he suggests, will become colonized by work and email. The episode also weighs whether autonomous cars will be safer (no definitive proof yet exists), how vehicle design will shift toward cute, unthreatening forms, and why the trolley problem misses the stranger dilemmas—whether a machine has a duty to act heroically—that Asimov's Laws of Robotics frames more clearly.
Torchinsky argues that the central challenges of autonomous cars are human, not technical: semi-autonomy is fundamentally incompatible with how humans operate, fully autonomous vehicles will stop being 'cars' and become reconfigurable mobile rooms, and society risks losing the meditative, productivity-free experience of driving itself.
- Level-two semi-autonomy demands sustained human alertness that humans cannot deliver, making it more dangerous than either full driving or full autonomy.
- Once a vehicle drives itself it becomes a transportation robot designed from the inside out, not a forward-facing car.
- Human driving carries irreplaceable value as a refuge from productivity and as the experience of a journey, which an autonomous future erases.
People will demand self-serving settings over collective benefit, making safety-first programming politically impossible.
- Given how poorly people accept minor collective sacrifices — illustrated by widespread refusal to wear masks during a pandemic — it is unrealistic to expect drivers to surrender their car's autonomy or accept their property being sacrificed for the greater good; many would choose settings that prioritize their own survival over surrounding people.
“We live in a country where people have trouble wearing masks in the middle of a global pandemic, so I'm not at all optimistic they're gonna turn over the autonomy of their cars to companies.”
Trolley problem scenarios are too rare and philosophically misframed to guide autonomous vehicle ethics.
- The trolley problem is largely misapplied to self-driving cars because it was meant to highlight competing ethical instincts rather than teach how to act, and such kill-one-versus-five scenarios are very rare; the more genuinely difficult and freaky questions are Asimov-style ones, like whether an autonomous vehicle has a positive duty to act heroically to stop a terrorist.
“The trolley problem is not supposed to teach you something about ethics. It's just supposed to remind you that we have competing ethical instincts and it's not clear how to resolve them. And I think that it's almost misapplied for self-driving cars”