Hacker News new | ask | show | jobs
by qeternity 3536 days ago
The bigger issue is that autonomous cars make it extremely easy to massively boost utilization: driver assist still means your car sits at home, work, etc when not in use. This in turn means more cars are required by society. Autonomous cars can achieve much higher levels of utilization.

This is the real reason legacy car makers don't want fully autonomous vehicles.

6 comments

A car being driven all day (especially in cities, where the demand will be highest) is going to wear cars down incredibly fast. Manufacturers won't see that big of a hit if fewer people buy cars, if those people will replace it every year or two.
If it is profitable, it will be financeable.
I don't think the incentives quite work that way. When cars can be utilized at ~90% instead of >10% as they are now, that will make them useful for people who didn't previously use cars. Senior citizens who are too old to drive themselves, people who are too scared to drive or too bad at it, families that can only afford one used car, city dwellers who don't have parking spaces, probably more.
The demand for repair parts will increase significantly and cars will expire a lot faster.

Cars that do 400 miles a day will reach 200'000 miles within ~1.3 years. Usually most cars are fairly low value by then. They would reach 400 miles by being on a highway for ~6.5 hours/day, which is fairly reasonable to assume.

I think autonomous vehicles will be priced taking that into consideration. It is just expectation of people that they will be priced as cars and can be deployed as profit generating asset.
I'm not sure though. There would be the same or more total miles driven. So they'd have to replace the cars that much faster.

Tldr: It should be a wash.

There's a lot of steel in today's cars that won't be needed if they're optimized for moving commuters around cities one or two at a time.
Even more than that, if cars don't crash, they don't have to be crashworthy. BMW's concept motorcycle concept makes that point: If autonomous vehicles mean that road crashes become very rare, as other rare as, say, accidental strangulation in bed, you won't need a helmet to ride. A vehicle could be an minimal as a motorcycle and still be much safer than an S-class Mercedes.
I doubt safety will ever get to the point where crash worthiness is not an issue. The Washington DC Metrorail system is an excellent example. It was built to be fully automated, with the operator being there only to close the doors at stations.

The systems have suffered many failures and crashes over the years, due mostly to poor maintenance and gross mismanagement.

Now the agency is under pressure from government regulators to retire old cars because they are not crash worthy. I find it alarming that we would talk about rail cars being crash worthy. Why not improve the system so cars don't crash? Nonetheless, new cars are much stronger and more crash worthy.

Since there are many more possible points of failure with autonomous cars (bad roads, non-autonomous cars being out there too, objects in the road, bad programming like that involved in the Tesla crash, etc) autonomous cars are going to have to be at least as crash worthy as Metro cars for the foreseeable future.

Maybe. I'm not aware of much precedent for the idea that statistically safer X leads to fewer safety regs for X. Maybe you could draw a parallel with the over-the-counterization of various pharmaceuticals, but I'm not sure that's an analogy that would fill me with hope.
High speed rail relies on active safety. The carriages do not meet the crashworthiness standards of conventional rail.
Interesting! I didn't know that.
Wouldn't Intel not want "the cloud" for the same reason?
My guess is that the cloud doesn't affect utilization that much. Auto scaling (IMO) is fickle and for the most part the same number of servers are running, in fact probably way more servers are running because their easy to provision (and then forget about).