Self-driving cars can optimize limited road infrastructure. Of course, this means ALL cars would have to be in on it, but that isn't infeasible for 10 or 20 years from now.

So I kind of agree with parent. Doing something today might be outdated given tomorrow's technology enabling better solutions. But that is always a risk in infrastructure investment.

Why not just never improve anything, ever? I mean we might have self-driving cars in 20 years, but another 20 after that there will be something even better than self-driving cars, and who wants to own something that will be obsolete in 20 years, anyway?

We will have self driving cars in 3 to 5 years at this point...the writing is on the wall. We will have pervasive self driving cars in 10 or 20 years.

So to ignore something that is pretty much a sure bet to redesign your city over 20 years, when what you are building is meant to last 50-100 years, is kind of stupid. You do what you can with all information available.

I think you're overly optimistic with that estimate -- we'll have "driver assist" cars in 3 - 5 years that can navigate most roads in normal conditions, but will still need the driver to step in from time to time to help with unexpected conditions.

10 - 20 years is more likely for true self-driving cars that can operate completely without human control.

Traffic congestion is also a question of person density. If there were ways to increase people density per sq ft on those roads, with buses or, heck, some weird quadruple-stacked short-distance car transports, you can reduce traffic and commute loads as well.

Traffic is a compression wave. You brake, then the person behind you brakes, then the person behind them, etc. You accelerate, then the person behind you...

Highways that become automated-only will not have this problem. The cars can communicate with each other and change speed in unison. You don't even need to convert a whole highway, you can just reserve a single lane for automated cars and get most of the same benefits.

The near gridlock on SOMA streets every rush hour isn't because of a compression wave - there are simply too many people trying to drive out of the city at once. Even if magical cars could erase congestion on freeways, there's still the problem of not enough parking for everyone that would have to drive without public transit.

And don't say "But self driving cars will fix everything! They don't even need to park, they can drop you off at work and drive away". If the cars aren't parked, they are contributing to congestion. And since commutes aren't generally balanced, it's not like the self-driving car will leave Walnut Creek at 8am, drop you off in SF at 9am and then pick up a new passenger to head back to the East Bay - that car is going to be loitering around SF somewhere, waiting to take someone back home to the East Bay.

SF Muni carries 600,000 passengers/day. BART carries 400K (though not all to SF). Caltrain contributes another 50K. How could you possible accomodate all of these people without transit?

Like I said, building effective transit in existing cities is very expensive -- the BART system cost around $1.5B when it was built (around $20B today), but few would argue that SF would be better off without it.

>If the cars aren't parked, they are contributing to congestion. And since commutes aren't generally balanced[...] that car is going to be loitering around SF somewhere

You're right, traffic patterns are not balanced, so the cars returning to the hot areas for pickups will not be contributing to the traffic in the most-congested direction.

Also, SF already has BART, Caltrain, and SF Muni. I'm not familiar with the specific history here, but transit and housing are the limiting factors for where people live in the first place, so the number of commuters will expand when capacity increases. The better question is, what if SF didn't build those systems? It wouldn't be San Francisco today minus the trains, it'd be a totally different San Francisco. So you can't really take the ridership numbers from today as evidence that city couldn't cope without trains.

Also, this was all happening when self-driving cars were not around the corner. It's a different story today.

>You're right, traffic patterns are not balanced, so the cars returning to the hot areas for pickups will not be contributing to the traffic in the most-congested direction.

Except that in a city, all traffic contributes to congestion,there is no single "commute direction".

>Also, this was all happening when self-driving cars were not around the corner. It's a different story today.

That "corner" is decades away, meanwhile, people need to get to work today -- and you still haven't explained how self-driving cars can replace a million+ trips/day on city streets that are already at capacity during rush hour. They may be self-driving, but they still drive on roads.