Self-Driving Cars Might Finally Be For Real This Time

Episode Summary

Executive Summary: The episode argues that self-driving cars were overhyped a decade ago but are now re-emerging as a real, gradual commercial product, led by Waymo and Cruise. Guest Tim Lee says the hard part is not basic autonomy but the endless long-tail of rare edge cases, plus safety, operations, regulation, and economics. He is optimistic that robotaxis will eventually be profitable and widely adopted, though likely first in sunny, car-dependent cities rather than New York.

Main Topics: The return of self-driving car optimism (Priority: 5/5): The hosts revisit the collapse of the old hype cycle and ask whether autonomous vehicles are finally becoming practical after years of disappointment. Why the last 5% is so hard (Priority: 5/5): Tim Lee explains that the core challenge is not routine driving but rare, unpredictable edge cases such as fire scenes, police tape, wet concrete, or unusual road conditions. Industry structure and surviving players (Priority: 4/5): The conversation maps the current market: Waymo and Cruise as leaders, Tesla as a driver-assistance company rather than true robotaxi operator, plus smaller players like Zoox, Motional, Mobileye, and Aurora. Business model and economics of robotaxis (Priority: 5/5): The episode explores why removing human labor could make autonomous taxis economically attractive, while noting the need for expensive sensors, remote operations, and sustained capital investment. Safety, regulation, and liability (Priority: 4/5): The guests discuss how self-driving systems must be near-perfect before launch because failures can cause injury or death, and how regulation varies by state. Potential effects on cities and transportation (Priority: 4/5): The discussion looks at broader impacts such as less parking demand, new congestion pricing models, package delivery uses, and shifts in urban planning.

Key Arguments: Self-driving cars failed to live up to early hype because the remaining problems are rare but consequential edge cases that require enormous effort to solve. Autonomy is less about breakthroughs than about years of iterative grinding through unusual scenarios and operational refinement. The main bottleneck is more software/data than hardware; cameras, radar, and LiDAR already exist, but the systems need more examples of rare situations. Waymo and Cruise appear to be moving from testing to real commercial service, suggesting the technology is closer to a viable product than many assume. Tesla should be understood as offering driver assistance, not a true unsupervised robotaxi system. The economics could work because human drivers are roughly half the cost of ride-hailing, so removing them could make robotaxis profitable if sensor and ops costs fall. Deployment is likely to expand first in easier, sunny, car-oriented cities before dense, weather-challenging markets like Manhattan. Urban and logistics applications beyond passenger taxis are plausible, but will likely require specialized vehicles and business models.

Data Points: Podcast format: 5 minutes or less - Stock Movers promotional ad describing short Bloomberg audio updates Waymo driverless Phoenix service start: fall of 2020 - Tim Lee says Waymo began fully driverless operations in Phoenix then California approval for San Francisco commercial service: a week or two ago - Tim Lee notes regulators recently approved Waymo to begin commercial operations in San Francisco Waymo footprint growth target: 10x by the end of next year - A company target mentioned by Tim Lee Cruise revenue target: $1 billion by 2025 - Tim Lee cites Cruise's stated goal, implying a roughly 50x increase Human road fatalities: 30,000 to 40,000 per year - Tim Lee contrasts self-driving safety concerns with existing human-driven crash deaths Human driving exposure: billions or trillions of miles annually - Used to explain why self-driving car safety comparisons are still statistically limited Fatality rate reference: about 1 death every 100 million miles - Tim Lee's rough benchmark for human driving risk Self-driving miles driven: tens of millions of miles - Explains why the sample size is still too small for definitive safety conclusions Waymo/Cruise deaths so far: zero - Tim Lee says the leaders have had no fatalities despite substantial testing Uber fatality: 1 pedestrian killed in Tempe, Arizona in 2018 - Tim Lee says this effectively destroyed Uber's self-driving program Safety-driver period: first few years of testing - Waymo and Cruise had safety drivers in cars before removing them Phoenix service territory: a couple hundred square miles - Current limited operating area for Waymo's service

Pivotal Quotes: "The death of self-driving cars is greatly exaggerated." — Tim Lee: Title of his recent piece and the central thesis of the interview "It’s just a lot of little things. I mean, that’s the thing about a long tail is there’s a lot of stuff out in the long tail." — Tim Lee: Explaining why edge cases make autonomy difficult "I think in the long run, the taxi industry will be operated by self-driving cars." — Tim Lee: His long-term bullish view on the business model

Implications: Robotaxis may be closer than skeptics think, but progress will be slow, city-specific, and capital-intensive. If costs fall, the biggest winners could be ride-hailing, urban planners, and fleets; losers may include drivers, parking-heavy cities, and firms that miss the operational grind needed for scale.