If buying the hardware is even an option to begin with, there's going to be a hefty fee to subscribe to the required 'cloud based' self-driving service.
(And there's certainly no way individuals/small businesses will be able to buy such a car and run a self-driving taxi service profitably when a megacorporation could have all the income from all the cars)
Positive HN discussion: https://news.ycombinator.com/item?id=36927971
I think the argument is that we won't see the same availability beyond level 3, since level 4 is the first step where a human is not required behind the wheel. This means you probably need subscription to have a remote human who can step in and take control in edge cases, at least until we develop AGI. Even then, I suspect the economics of running a full AGI from each vehicles will make a subscription mandatory.
But I believe regulatory capture and anti-competitive tariffs are likely to make affordable, buy-and-use cars impossible to sell in the US. Regulations will likely change annually, like they do for cars right now. The govt will be worried that China has control of our morning commute.
I hate being cynical, but cars do seem like a lost cause.
But the reality is that people do own cars, and no car manufacturer that I know is also trying to become a taxi operator (except maybe in the indirect investment sense). Or in other words: sales in a market aren't affected by the abstract possibility of vertical integration, because any entity that attempts to vertically integrate will be undercut by other parties making sales at the same time.
Car makers operating according to this incentive will try to discourage taxi-like functions in self-driving cars, perhaps by only allowing operation when specified people are recognized by the car (anti-theft feature!).
Cars were hardware. But now contain ever-increasing amounts of software. With self-driving cars, the software is the main feature.
The development cost of the updates and patches has to come from somewhere - either significantly marking up the initial purchase cost to fund the on going development (side bit: when a one time purchase server shuts down people are up in arms about that) or with some form of subscription model.
If a driverless car maker can be liable for the software that they released some time ago and didn't update, then the funding of that development effort has to come from somewhere.
If some of the compute or data updates is in the cloud, that comes with an ongoing cost. Compare with getting a DVD with a database of road updates once a year ( https://ford.navigation.com/product/Catalog/2023-Green-DVD-M... ) and the "this road is closed" ( https://www.wyoroad.info/pls/Browse/WRR.STATIC5?SelectedDist... ) or "there is heavy traffic, a different route is suggested" services. Those come with an ongoing cost somewhere.
SaaS is the plague on software development, because the subscription model demands a shift to constant feature "upgrades" and other frequently-useless "updates" intended to justify the perpetual subscription costs to customers.
Even well before some such future, the type of people who buy Bugatti Veyron will still want to buy something unique and out of reach of the 99.99% of the population. Maybe that is human driven but what if human driven becomes illegal?