Hacker News Reader

Comment by paul | Hacker News Reader

The good news is that by the time they are finished, the whole thing will be obsolete. Self-driving technology is going to completely change urban transportation.

View on HN

I will put $100 into an index fund and bet the proceeds that self-driving technology will not be competitive with rail by 2040. The technology will be much harder to bring up to the necessary level than anyone expects, and without banning human drivers (politically impossible for the foreseeable future), you'll never be able to match the capacity of rail.

Even with perfect software how could cars, self-driven or otherwise, ever compete with rail? A four-track rail system has a per-track throughput of 100000 people per hour. Freeways can get 2000 cars per hour per lane, and an average of 1.6 people per car works out to only 3200 people per hour. That's the theoretical best. Actual roadways never approach this figure.

They aren't mutually exclusive. An always-available self-driving car service would drastically help the usability of subways or commuter rail in the crucial last 2-3 miles. If I could hail a car to drive me the 2 miles to the Caltrain stop and then hail another one to drive me the 3-4 miles from 22nd St. to the Sunset, a much greater percentage of the city becomes available to me without a highway trip.

I'm interested to dig into this question a bit more -- do you have any linkable resources that give more detail on those numbers? Thanks.

> Freeways can get 2000 cars per hour per lane

I'm missing something, in a single lane at 60mph you would be able to get ((60 * 1608)/5) (assuming car is 5m long) = 19296.

If we assume that there is 5m of seperation between every car (possible when all cars are automatically driven and communicate with each other) that's still 10k cars an hour.

4 lane highway, 40k an hour, assuming occupancy at 1.6 that's 64000 an hour.

Human driven no way, we are too unpredictable and react far too slowly to safely drive in that way but computers probably could.

Well you just made a pretty unfair comparison since you added up to 64000 for all four lanes vs 100000 people /per track/ on a 4-track railway. But I think your assumptions are all pretty optimistic. You wouldn't be able to run cars that close together on a multilane roadway because cars would need to move between lanes to enter and exit which would cause pipeline shocks if there was that little room.

Anyway the main difference between my claims and yours is that yours are speculation, while mine are based on history. We've seen five million people disembark at actual railroad stations. It's happened. We don't have to speculate.

> I will put $100 into an index fund and bet the proceeds > that self-driving technology will not be competitive with > rail by 2040.

Longbets? [1]

(They don't seem to be very active though, so if anyone knows better alternatives, you're welcome to share)

[1] http://longbets.org/

That idea is utter nonsense.

Self driving technology won't allow you to fit more than four people in a four person car.

Self driving technology gets you maybe a 4x-10x increase in the number of people transported per road lane by ensuring cars are mostly full, having them close together and cutting down on trips to reposition vehicles... however, that still won't be enough scale of transportation for the cities of the future.

It's a huge social/economic issue though. The people who can afford self-driving cars in our lifetime are not (usually) the same people taking the subway to work every day.

Reply on news.ycombinator.com