Self-driving cars have lately captured the imagination because they're futuristic — and because the last big futuristic story, electric cars, hasn't panned out.
You could argue that some of the enthusiasm around EVs from a few years back has eroded because a self-driving car doesn't require an alternative propulsion platform or several hours to recharge. You can rig up a normal gas car to drive itself.
Autonomous vehicles, logically, could fall victim to the same type of story change. The core vulnerability is a major shift in how we organize urban life, moving away from cars and even mass-transit to walking, biking, and more compact means of getting around. (The growth of cities in the 21st century and the change in how suburbs have their transportation organized will be a big threat to traditional mobility systems, such as automobiles.)
Conventional commuting — in some metropolitan areas a reliable productivity destroyer — could also be toppled by more advanced forms of tech-enabled telecommuting, a 21st-century megatrend that keeps banging up again the mid-20th-century vestige of consolidating workers in one physical place.
At the moment, Tesla has the best commercialized self-driving technology in Autopilot, but it really only works optimally on highways.
Newer arrivals, such as Cadillac's Super Cruise, are designed primarily for highway use. Through its acquisition of Cruise Automation, Caddy's parent General Motors is rolling out a fleet of self-driving Chevy Bolt EVs, but they're likely to be used in a "geo-fenced" urban setting.
We've had cruise control for decades in cars, and "adaptive" cruise control for about ten years (cruise control that uses sensors to vary speed based on traffic). If people use it at all, they use it on the freeway. And a lot of drivers don't bother because traffic compels then to switch cruise control on and off.
Self-driving cars could be OK for stop-and-go traffic and highway cruising, but the big question is whether limited use cases will discourage drivers from taking the plunge on the tech.
Many auto-industry folks think that fleet applications for self-driving tech make the most sense, but we won't know if the economics are better than old models (taxis, car services) until we see some numbers.
Any technology that can currently serve up high-level autonomy is going to be witheringly expensive. Tesla is trying to hold costs down by using cameras and sensors instead of pricey Lidar (laser radar) systems, as well as machine learning through its networked fleet of cars.
But thus far the only truly impressive self-driving tech has come from Waymo and the vehicle formerly known as the Google car — a Lidar-based approach.
It might be possible to add functionality to less sophisticated systems, but in that case, full autonomy would be elusive.
Regardless, anything beyond advanced cruise control is going to be costly. Newer systems will add thousands to the purchase price of a vehicle, unless the automaker decides to eat the expense, something that could be feasible with high-profit-margin luxury vehicles.
The expectation is that the price of the tech will come down with wider adoption, just as it did with antilock braking and airbags decades ago. But for that to happen, consumers have to start demanding the tech.