Hacker News Reader

Comment by edent | Hacker News Reader

The thing I don't get is... how are "self driving" cars better than public transport?

The two advantages I see is that you have a private space. And that they are point-to-point.

Those are important advantages. As, I suppose, is the fringe benefit of finding work for mathematicians.

But it seems to me you could do "self-driving" a lot cheaper and easier by improving the commons. Embedding sensors in the roads, street signs, and furniture - isn't that going to make more sense than turning ever vehicle into a super-computer?

Even after all that... buses and trams and light rail work absurdly well, at a fraction of the cost.

View on HN

In addition to the two advantages you’ve already covered, Americans regularly travel between points without sufficient density to support frequent and convenient public transit.

There's no reason you can't have frequent and convenient public transit as a subsidized service. This also ignores how between the years of 1938 and 1950 National City Lines — with investment from GM, Firestone Tire, Standard Oil of California (through a subsidiary), Federal Engineering, Phillips Petroleum, and Mack Trucks — acquired and dismantled the streetcar rail transit systems of 25 cities, including Los Angeles. We have worse transit systems in the US today than we did a century ago. All in the service of cars which we rebuilt our cities around.

While we fucked around for a decade the CPC built out a high speed rail network on the mainland. Similar landmass, similar geographical challenges, less wealthy nation. Guess Americans will have to wait until public transit is more profitable than making everyone get their own car with its very own JARVIS.

You really can't have convenient public transit that replaces cars in our current carcentric metroareas. The density isn't there. You'd have to totally dismantle the suburbs, which most people live in. Unless you mean point-to-point buses, but that's just uber.

You could slowly do it over decades in places primed for growth. Places like Arlington, Va are being urbanized.

You are underestimating just how empty large swaths of the US are. As someone who's driven cross country a couple times, its just too vast for it to make sense in a lot of places.

I'd imagine "Autonomous Bus that runs the same limited route every day" is a far lower bar to clear than "Fully Self-Autonomous Car" and would make a huge amount of public transit feasible in areas where it currently isn't.

Just go talk to a bus driver for a while about how absurd that claim is. "Same route every day" doesn't mean "It's the same things every day." You see some absurd stuff after a few years of driving a bus.

That argument can also be extended to trains. Problem is that the unions just wouldn’t allow it.

I’d love automation of public transportation, but nobody wants to try because of the social hurdles.

There’s already extensive road infrastructure pretty much everywhere. Autonomous cars that could just use that would be much better than having to build out and maintain rails.

Buses do not work absurdly well. They take three times as long to go the same distance and operate on a fixed and often infrequent schedule. They aren’t available all the time. You can’t easily transport cargo in them.

Ideally we’d redesign cities to be more walkable and bikeable so cars aren’t as necessary, but that’s not going to happen overnight.

Buses in the USA with limited funding do not work absurdly well, no. I've ridden buses in Hong Kong that are very nearly as efficient as anything else, and one comes along very frequently. Some of that is that Hong Kong is small, but it illustrates the point that buses can be better if funded well.

Spend as much money on buses as the Boring Company wastes with silly unusable routes nobody wants, and you'll get great bus service.

I actually live in Hong Kong and don't have a car and that has nothing to do with buses. Everything is within walking distance. Within 5 blocks of my apartment off the top of my head I can think of 4-5 grocery stores, 3-4 butcher shops, a wet market, 3-4 vegetable stands, 3-4 different dentist offices, several different daycares / pre-schools, at least 2 gyms, a public pool, about a dozen hair salons / barbers, dozens of restaurants and bars, and on and on. There is literally nothing I need on a daily basis that I couldn't walk to. I almost never take the bus anywhere.

If you had to commute regularly from, say the New Territories in to Central (about as far as you can go in HK), the bus/MTR takes 2-3 times as long as driving.

I am not suggesting that the bus service is essentially for daily life in HK, just that it is remarkably efficient for getting from Sai Wan Ho to Central and back again, for example. It's efficiency would amaze more US residents, who've never seen a bus system so well-funded and well-operated.

FWIW, Sai Wan Ho to the New Territories took me MTR and light rail, where I was ultimately picked up in a car by the person I was visiting. People living in the New Territories seem to find a car nearly-essential, but that seems to be because MTR doesn't run out there, and the bus service is not nearly as well-developed as Hong Kong Island or Kowloon.

> Autonomous cars that could just use that would be much better than having to build out and maintain rails.

Not if everyone has to spend $50-100K on a car + maintenance + insurance + fuel + parking every 10 years or so and live in cities optimized for those. Regardless how much it costs to invest in infrastructure it will cost far less than equipping everyone in America with a new car at their own expense and socializing the cost of maintaining roads, highways and interstates.

Replacing each car in America today with an electric, self-driving car (using a Tesla ballpark and factoring in the aforementioned additional expenses) is $10-20 trillion dollars. Imagine the kind of public transit we could have for twenty trillion dollars invested every 10 years. China put in 24,000 miles of high-speed rail for $500B.

People don't have to all run out at once and buy a new car right away. They'll just replace their current car whenever they were going to anyway. Prices will come down over time. Taxi services that don't have to pay drivers will be far more abundant and affordable in the meantime.

Spending that $10-$20 trillion by the government would mean collecting it all at once from people via taxes, not them spending it at their own discretion. That's a big difference.

I didn't propose investing $10T at once in transit, I'm saying people are currently spending about $1T/yr on cars and that could be spent on a stellar public transit system yielding much more livable cities.

Remember, America is spending money hand over fist on the socialized roads, freeways and interstate all so people can then spend a ton of private money on a car averaging 1.5 person occupancy that spends 95% of its time sitting still not doing anything.

But every person who is buying a new car in a given year (usually via a loan they pay off over 3-5 years) is not going to just give that money to the government to build a train line that might be finished in 10 years instead (and they have no transportation in the meantime). That makes no sense. These things are not interchangeable.

This is a well-trodden path. The government finances the projects over a number of years through the capital markets and then levees a tax to cover the payments. They are quite interchangeable. 30% of New Yorkers and 35% of DC residents don't even have access to cars. 90% of Hong Kongers don't have cars. Yet they manage to thrive.

It's not replacing the money that gets spent on cars. People buying a new car are buying it because they need it now, they need to get around today, not in 10 years when the new train gets built. They're still going to buy the car and then have to pay the tax to build the transit on top of that.

So you've identified the chicken and the egg, I agree, which is why we finance these things. By the time its built that money won't need to be spent on cars anymore on an ongoing basis.

I'm all for public transport, but lying to people and saying it's going to be zero cost because it will completely supplant what they spend on cars is not the way to get support for it. A train or bus line that goes to 20 specific places and nowhere else is not a substitute for a car that can go directly to any place, whether across town or across the country.

We should absolutely invest in public transport, but let's not pretend it isn't going to take making spending money on it a priority. It doesn't come for free and it doesn't completely eliminate the need for cars.

It also requires the entire development to shift to be much, much denser. It's not like "oh we just issue some bonds to build a rail line and then nobody ever has to buy a car again".

Public transportation is not a complete substitute for cars. You still need roads.

Self driving has the potential to provide a middle path. Vehicles with 95% up time instead of down time.

The reason that cars spent 95% of their time unoccupied is because most people only need to travel during certain times - the same times for everyone. That’s really all there is to it. As a result you need capacity for peak at all times. It can’t be made more efficient except by train, bike and bus.

This is what Uber and Lyft claimed they could solve over the last 10 years and made zero progress.

Self-driving cars solve the last-mile problem of public transport. They also solve the problem of wait times at bus stations, and while buses stop along the route. In all you'd do most trips in about half the time in self driving cars than on a bus.

In theory the economics would work out so you wouldn't want to own a car anymore, just use these robo-taxis, as it would be cheaper and more convenient. We'll see if it really plays out that way, and how long it takes.

> In all you'd do most trips in about half the time in self driving cars than on a bus.

Does that assume there is no traffic? I feel like the space savings from buses (in terms of space on the road) ought to mean that they would be much quicker than cars for most trips if they are given priority as they'd eliminate traffic jams, significantly increasing the road speed. This applies only to dense urban areas of course. For rural areas, self-driving cars make quite a bit of sense.

Unless you're in an apartment with a bus stop right outside, the losses from walking to the stop, waiting at the stop, having the bus stop all along it's route, and then walking from the final stop to your destination will usually dominate compared with traffic. Even in cities with dedicated bus lanes.

In theory you could save time in special circumstances on a bus, but I don't think I've ever done it in practice. One exception could be if you need to search for parking at your destination. A robo-taxi wouldn't have that short-coming.

This is not true at all. Unless you're going very close by, I usually spend longer in the bus/train than walking or waiting for it.

Good public transit cones every 15 minutes, so on average you'd be waiting at the stop for 7.5 minutes. Most bus rides are longer than that

Once you get out in the suburbs it's every 30-60 minutes. You've usually got at least 5 minutes walking at each end. Sometimes in inclement weather. There's always exceptions, but there's a reason people use cars when they can afford it.

Trains on the other hand can be a lot better if the stations are close to where you live and where you're going, they don't have traffic and they tend to come every 5 minutes or less during peak hours.

Give them all the priority you want to, they still have to make 15 - 20 stops along their route and that is why the bus takes so long. You try going 10km and stopping every few km for 2m at a time and see what it does to your transit time.

Busses aren't faster.

Some cities have started implementing Bus Rapid Transit [0] for these longer journeys, which are buses with fewer stops and a requirement that you buy the ticket before you board (or use contactless payment). This gets you lots of the travel-time advantages of rail without having to build out the infrastructure. These can interchange with slower, stopping buses to provide a comprehensive transit network.

[0]: https://en.wikipedia.org/wiki/Bus_rapid_transit

FWIW- California has spent over 70B on a single train line from San Francisco to Los Angeles. It hasn’t even been half finished. In addition the US passed a bill to spend over 1 trillion on infrastructure. Taken together, I don’t think that improving the commons will be cheaper than $100 billion.

In fact, as someone else mentioned, being able to utilize the existing infrastructure in new ways will likely be cheaper.

White Elephant.

Just ask disabled or elderly people and the benefits will become more clear. Basically anybody that can use a normal taxi, but at a lower price (presumably though not proven yet)

Hey you just made clear to me what bugs me about self driving cars. They make me feel disabled.

Public transport is great, but self-driving promises to fix the last mile problem.

> But it seems to me you could do "self-driving" a lot cheaper and easier by improving the commons. Embedding sensors in the roads, street signs, and furniture - isn't that going to make more sense than turning ever vehicle into a super-computer?

This would require a broad agreement on a global standard for Infrastructure to Vehicle (I2V) and huge investments by local and national governments around the world. It could technically happen, but likely will proceed at a snails pace. As you mention, it would be much better than relying on-board sensors like Lidar and cameras as centimeter accurate GPS coordinates can be embedded into road signs etc, but it will take decades.

US is extremely car dependant while being the biggest economy in the world. I don't think there is any other reason.

The logical thing would be to make cities less car-dependant but I'm sure there is strong economic interests that would like to continue selling cars.

Trains have been self-driving since... 1981?

There's really no advantage to self-driving cars or electric cars. They're dramatically less space-efficient, they're incredibly expensive (car + gas/electric + maintenance + insurance + parking + time spent trying to find parking), highways ruin cities and even car-first streets hurt communities. They promote unhealthy cities and lifestyles. Also, they're super duper complex in comparison.

The optimal solution is rail, followed by bikes and e-bikes then by busses.

> The optimal solution is rail, followed by bikes and e-bikes then by busses.

The "optimal solution" to the housing problem would be to have everyone live in dormitories and eat in communal chow halls. For extra efficiency, you could hot-bunk the beds three shifts per day.

But, oddly, no one wants to do that.

There is more to life than "efficiency".

What a silly comparison. You get what you pay for and if you under-fund transit you end up with bunks. If you fund it appropriately you get gorgeous and efficient systems like Hong Kong MTR and Singapore MRT. Or even the new electric Caltrains [1].

[1] https://twitter.com/Caltrain/status/1559318073940647936/phot...

> What a silly comparison.

What's silly is suggesting that "efficiency" is the only metric for quality of life.

> Hong Kong MTR and Singapore MRT.

Hong Kong has a population density of 6,801/km^2. Singapore has a population density of 7,804/km^2.

The United States has a population density of 33.6/km^2. It turns out that this matters.

The parts of the United States that have high population densities (though still not nearly as high as Hong Kong or Singapore, in general), such as the large East Coast cities, Chicagoland, etc. already have public transportation.

> What's silly is suggesting that "efficiency" is the only metric for quality of life.

That's a straw man you introduced. Cities with fewer cars are more efficient and have a higher quality of life.

That comparison re: density is meaningless, and that number doesn't matter at all. Nobody's going to put Hong Kong type subways in the middle of the Nevada desert where that average gets pulled down. Only New York and Chicago really have proper transit and that's only downtown. There's no reason for car ownership to be higher in SF or NY or DC or Chicago than in Hong Kong except inadequacy of transit.

You'll notice that all the most liveable cities in the world have excellent transit and none of them are in America [1]

[1] https://www.visualcapitalist.com/ranked-the-most-and-least-l...

> Cities with fewer cars are more efficient and have a higher quality of life.

Sure thing. That's why everyone who can possibly manage it moves to the suburbs when they have kids. Because they're looking for "lower quality of life" and a less "livable" environment for their kids.

Note that Europeans do this too.

P.S. I can assure you that Calgary, #3 on this list, has plenty of cars.

> There's no reason for car ownership to be higher in SF or NY or DC or Chicago than in Hong Kong except inadequacy of transit.

Except that people enjoy having a personal vehicle that goes where they want, when they want, and also enjoy not dodging junkies and drunks pissing in the subway cars and shaking them down for money.

It's more of a mixed bag in my book:

- On the one hand, "self-driving cars" is a general problem of transporting anything from point A to point B without direct human intervention in a very complex space, we'll probably see a lot of things getting transported with cheap point to point solutions post the tech being widely available with some people aggregating pickups and drop offs for efficiency a la trains etc.

- On the other hand, I expect a wildly high spike in traffic post these vehicles being more accessible, currently the largest limiting factor to a car being on the road is someone driving it, what happens exactly when practically everyone who can afford a car now can just send them off to do their daily todo list stuff like picking up groceries. There was a fun study[0] that showed car usage increased[0] on average.

I suspect peoples opinions may shift as roads become more train-like with ever growing slow moving traffic of either empty cars doing tasks or people reading as they slow travel to their destination.

Trains / buses on bus only routes / lanes restricted from being self-driving / premium lanes may be the only fast ways of getting around.

- [0]: http://www.joanwalker.com/uploads/3/6/9/5/3695513/mustapha_e...

- [1]: "while total VMT decreased slightly during the chauffeur week for the first subject, and hardly changed (on average) for the second subject, the remaining 11 subjects increased their auto travel. The increases in total VMT during the chauffeur week ranged from a low of 4% for one of the Millennials (from 532 to 554 miles) to a high of 341% for one of the Retirees (from 117 to 516 miles), with an overall increase of 83% for the entire sample (from 3,344 to 6,118 miles)." where VMT is "vehicle-miles traveled".

From another perspective, if you assume that if you can make people more productive, your economy will get better, an issue I see when I travel is it's really hard in a lot of places to get from point A to point B. Take Mexico City, for example. I am not an expert but I spent quite a bit of time there and pretty much anything takes multiple hours of transportation unless its right in your area. And that's a problem for many people who cannot afford expensive means of transportation. If you could make it so travel times end-to-end were less than an hour and even shorter in more cases, people could accomplish more. I don't research this stuff, but I feels to me like a place that would have a huge ROI, and I just don't think self-driving cars improves that for a lot of people.

> Even after all that... buses and trams and light rail work absurdly well, at a fraction of the cost.

"Absurdly" is far from true. They work barely at all if your time is valuable to you.

I can take a train into a nearby CBD - avg 15m wait time for bus (to train station), 10m bus transit time, avg 7m wait at train station, similar waiting times at destination, all multiplied by 2 because I return at some point.

It's usually between 45m and 60m.

In about 15m I can drive there. I'd rather spend the extra time with my kid.

Public transportation is an option in places that have the population density to support it. That is a huge part of the USA, not "fringe".

> The thing I don't get is... how are "self driving" cars better than public transport?

Self-driving cars are neither better or worse. They do not exist in reality, only in our imagination.

Such things are not bound by mundane constraints. They can be orders of magnitude safer, they can solve congestion, be fast cheaper, not use parking, and even somehow solve climate change.

I often wonder if ALL the cars are driven by AI would it be safer as the cars would behave more predictably, provided the code was bug free.

"But it seems to me you could do "self-driving" a lot cheaper and easier by improving the commons. Embedding sensors in the roads, street signs, and furniture - isn't that going to make more sense than turning ever vehicle into a super-computer?"

I think this is the way to go. Could maybe improve safety which was is holding autonomous back.

Labor is a huge component of the cost of bus public transport. If you had a driverless bus, you could greatly expand system coverage and frequency and still probably save money.

There would still be challenges: how to handle safety issues, crashes, unruly passengers, etc. without a driver, but worst case you're paying for an attendant now instead of skilled labor.

I don't think there would be a labor savings. You are misusing the term skilled labor.

You don't think bus driving is skilled labor? Where I live, drivers earn about $45k with city benefits and require a CDL.

Where I live drivers make over 100k, and require many certifications, but that does not mean it is skilled labor.

Skilled labor usually refers to individuals who work in jobs that require experience, and education, like college degrees and advanced degrees, and who are highly trained professionals in a specific field.

This is in contrast to driving a bus, where anyone can walk in off the street, go through 1 to 2 weeks of training, and then pass a certification or licensing test

Skilled v. unskilled labor uses fairly classist terminology. But its it a term of art--not a moral judgment.

Bus drivers are unskilled labor.

They only make decent money since its a pubic union job. Good for them. But if the government contracted it out at the lowest bidder, they wouldn't be making that much.

Cars are better than public transport in the two dimensions that you listed, plus the additional "it's 15 seconds rather than 5-15 minutes of waiting time to start the ride". (This would stay true for personally-owned self-driving cars, but not for hailed/shared self-driving cars.)

What would happen is multiple private companies would be trying to install their proprietary street tech aids all over the place.

Much like how the NYC subway now has different gauge tracks after all the competing companies were merged. Or the multiple cell networks we have now.

Obviously, good public transportation across the entire US would alleviate the need for a lot of self driving. The reality is, that will never happen for anyone outside the immediate area of a large city. Embedding sensors everywhere suffers the same problem.

I think embedded sensors are easily fixed in high maintenance areas.

In Canada, as with a lot of the US, the roads are asphalt. These tend to get resurfaced once a decade.

All we need is a cheap rfid type device, with unique serial numbers, and they can just be dumped into the asphalt, at least a few per sq cm.

Just mixed in, no specific placement, as long as the density requirement is met.

This allows for redundancy too.

After the road is done, a mapping device can drive along slowly, and map sensor location to road location, and create a database.

This may seem like a lot of data, but you only need this data for the area you are in, and if we started today, it would be a few years before roads were created this way.

Memory and storage grow constantly.

My biggest concern is ists of some type. Anti-technologists, some radical *ist group, dragging some high em device behind them, to blow all the rfid devices.

But anyhow, road sensors aren't that hard, or even expensive. Not compared to standard resurfacing costs, and that is done frequently enough that natural repair cycles can get things embedded.

You made me curious: https://pubs.usgs.gov/fs/2006/3127/2006-3127.pdf

There is almost 4 million miles (6.5 million kilometers) of roads in the United States, with just over 45,000 miles [73,000 kilometers (km)] composing the Interstate Highway System (fig. 2) (Federal Highway Administration, 2004). The total distance of U.S. roads combined could circle the Earth at the equator 160 times; the interstate highways, almost two times.

The analysis shows that a total of approximately 1.5 billion metric tons (Gt) of aggregates, 35 million metric tons (Mt) of asphalt, 48 Mt of cement, and 6 Mt of steel is in place in inter- state highways. These estimates illustrate the large quantities of these materials used in the Interstate Highway System.

This doesn't translate exactly to how much surface is covered, but I know locally all the highways are concrete. Also, placing sensors in 4M miles of roads is no small task.

Interesting. There is a lot more ashpalt here, in Canada. I suspect it is due to frost, and "frost greaves", which can buckle and damage even concrete roads.

Asphalt is cheaper, and replaced more often, so degradation due to frost greaves and such is not as fiscally damaging.

But my entire point is, by placing sensors during normal resurfacing, it literally is a small task.

I think mapping the roads is a problem that doesn't exist and not the hurdle to automate driving. You could do the same thing with a line of paint on the road and we already have those in most places. The real challenge is everything else that is happening on and around the roads.

One major cost factor in public transport is the driver. Uber like dial-a-ride services never caught on as public transport option because of this. Replace the driver by a computer and even if the car costs twice as much it becomes an attractive solutions.

There are other applications outside of everyday personal needs. E.g. the trucking industry

Those should really just be trains except for the last mile no? Then there's the really cool stuff like Switzerland's fully underground autonomous cargo delivery network. [1] A freight train can move 1 ton of cargo 400-500 miles on a single gallon of diesel and of course there's a lot of prior art in the self driving train space.

[1] https://spectrum.ieee.org/cargo-sous-terrain

You could combine the both to have self-driving small buses that are not driving on fixed routes, but optimizing route and schedule based on customer needs.

Cars and car infrastructure are environmentally unsustainable.

Electric and self-driving cars are not the solution.

The word needs trains, trams and [e]bikes.

public transport works well in dense urban centers. It makes no sense in rural areas where frequency of trips is low and distances high. And simply saying 'tough for the bumpkins' doesn't really work because the cities need the same roads to bring in resources.

False dichotomy--do both.

Reply on news.ycombinator.com