This blog post has been making the rounds on my Twitter feed, from a university professor contemplating the possibility of high-speed rail in Massachusetts:

In fact, it’s much easier for me to imagine my semi-autonomous car speeding down the Mass Pike as part of a computer-controlled platoon than boarding a train in my little city and disembarking in a bigger one.

There’s something very odd about a world in which it’s easier to imagine a futuristic technology that doesn’t exist outside of lab tests than to envision expansion of a technology that’s in wide use around the world. How did we reach a state in America where highly speculative technologies, backed by private companies, are seen as a plausible future while routine, ordinary technologies backed by governments are seen as unrealistic and impossible?

I could quibble with the piece’s definition of “public good,” but it offers an interesting perspective. HSR advocates have long argued that rail is more efficient for moving people between 100-500 miles. And yet the systemic change involved in securing rights-of-way and altering land use patterns seems completely intractable compared to simply swapping driverless cars into our existing transportation ecosystem. We HSR advocates may someday find ourselves on the losing end of a Betamax-vs-VHS debate, where the technologically superior product fails in the marketplace despite being better.

And yet, fully taking advantage of driverless cars will itself require a disruptive, painful transition in affecting everything from safety to ownership to infrastructure. As Dave Roberts wrote a while back in Grist, we should think of driverless cars not as widgets, but as part of a larger system, one which could obviate car ownership altogether. The savings potential in terms of economic impact and human life is enormous.

That said, if we’re changing the larger system anyway, then maybe the prospects of rail don’t look so dim. After all, Google’s super-expensive proprietary radar systems are still way too costly for consumer use. Meanwhile, the supposedly intractable and sclerotic federal government is finally allowing modern, lighter trains on US tracks. Large, disruptive changes to the status quo will necessarily create new winners and losers; that’s true whether we’re talking high-speed trains or driverless cars.

This post has been updated for clarity.

77 Replies to “HSR, Driverless Cars, and Systemic Changes”

  1. The term “driverless car” is itself a lie. It’s not just that something is driving it, namely a computer program. Computers have trouble with things like heat, cold, wet, shock, dirt, and surprises- which are the essence of ground transportation.

    But the main problem is that the ultimate driver is whoever wrote the program. The result of any calculation by a computer depends upon the worst piece of information any human being gave it- including the mistakes due to vulnerabilities designed into the hardware and software by people.

    Maintained, adjusted, and repaired by people.

    Including tired people, angry people, discouraged people, underpaid people, and overworked people. Bossed by stupid people, greedy people, ignorant people, and distracted people. Anybody encountered anybody in the IT world like this lately?

    Read the headlines about the Affordable Care Act rollout. And then tell me why you would ever get into a so-called driverless car, or ever get into any vehicle on a highway where there’s a single one of them.

    Mark Dublin

    1. I’m not a fan of driverless cars, but I don’t personally find the argument that software is less safe than human drivers persuasive. People in cars are constantly distracted by things like:
      – talking to a passenger
      – playing with a cellphone
      – being tired
      – being drunk
      – bright sunlight
      These are things you would not have to worry about with well designed and tested software, and though its certainly possible for there to be bugs in software resulting in deaths, I don’t think its any less dangerous than the current situation with human drivers.

      That being said, even if we had a driverless car system that worked much like a permanent taxi service everywhere, it would still be much less efficient then mass transit at getting people around for the energy expended. Just because you no longer have to leave cars parked for hours on city streets doesn’t get you around the problem of very limited capacity on those streets with the majority of people essentially going the same direction.

      1. Google Car is just a robot taxi, the effect is the same as very robust, pervasive and cheap taxis (imagine a taxi that cost the same as car depreciation, say .25 cents a mile, anywhere). The subsidized taxi service I proposed for Puget Sound in 1993 would be the equivalent of Google Cars.

        You say it would be much less efficient at getting people around. Here we quibble, because I imagine you envision a dense vertical city whereas I am content with suburbs. The article is about a man who wants to live in a rural area and commute to a jobs center — but do it as if he were very close. So he is saying, again something I’ve often said, that given the existence of 300 mph trains, what is considered rural, like Inland Washington, is now in effect, suburban!

      2. “even if we had a driverless car system that worked much like a permanent taxi service everywhere, it would still be much less efficient then mass transit at getting people around for the energy expended.”

        But you would have succeeded in preventing a lot of people from sinking a lot of money into a personally owned vehicle which they use by default- they would be much more likely to make a choice to trade time for money and use mass transit for any given trip, or an autonomous van-bus that lies somewhere in-between a taxi and a bus in cost and convenience or any number of other options that will become possible. So the overall efficiency ought to go up.

      3. But we are seeing the reality of the last mile. Given our BUILT infrastructure is predominantly suburban, and that building density has so far been linked to higher and higher prices, suburbs and exurbs are going to remain a reality. Further, some people choose to live in less dense environments.

        We have found that it is too costly to provide “convenient” transit service to neighborhoods and efficiency analyst say humans must walk up and down hills to get to service and then service is always at risk of being cut. Yet, we want people to not use personal cars.

        Google Cars are one possible solution to removing the need to provide costly service to low density neighborhoods by providing a way to get people to frequent routes and trains. They don’t need to be garaged for most of the day and by utilizing a fleet of cars, the total number of cars on the road could be diminished.

        It is one possible future. Other futures might include magically breaking the developer/higher prices cycle (only a socialist dream could bring this) to build affordable housing density in urban cores.

    2. Mark, the difference is that once you refine a problem out of the software, it will never make that mistake again, while humans just keep doing the same old things wrong time after time after time.

      Robot cars will initially have issues and fail in unpredictable ways. But as they and their programmers gain experience, they will get very good, very fast.

      I enjoy driving and will regret not being able to do it anymore. Nevertheless the safety and speed gains that are possible if all cars become robotized are just too great to ignore.

      1. What will happen is that the robots will be great in everyday situations and will fail catastrophically when weird stuff happens.

        Result: city streets & roads will shut down completely for blizzards, except for those few people who still know how to drive manually. Same for torrential rains. Et cetera. Dirt roads will be the domain of manually driven cars, with the result that there will be zero robot cars in rural areas.

        I am a programmer. It’s not possible to robotize all the cars. Too many cars and trucks exist precisely *for* the corner cases, the weird stuff, the unpredictable, the *stuff which you couldn’t do with rails*.

        If the majority of cars and trucks go robotic…. then it’ll become even more obvious that it’s cheaper to just put that majority on rails.

    3. Mark, computers are already very good at handling adverse weather and traction conditions; I’m not talking about self-driving cars that haven’t been released to the public yet, I’m talking about ABS and traction-control systems, which are extremely common features these days and widely acknowledged to perform their functions better than human drivers.

      Computers are already very good at avoiding colliding with things. The last few years systems to warn drivers when they’re about to run into things have become common; I think Volvo is introducing a version that supposedly will stop you from running into people and other things you might be aimed at, at least at low speeds.

      Then, FWIW, as of a year ago (per Wikipedia) Google’s cars had driven about 300,000 miles in testing with only two collisions. One was being driven manually around a parking lot; the other was rear-ended at a stoplight. That’s a pretty good record.

      Where you might expect driverless cars to struggle is reading and interpreting things created by humans: road signs and markings (especially temporary, non-standard ones put up during construction), behavior of other drivers, people directing traffic, etc. A lot of times human drivers struggle with unexpected road markings, too. Last night I took a bike ride on the new Broadway cycletrack and the half-finished restriping of E Green Lake Way and watched several human drivers fail to follow the new markings correctly.

      1. We’re not talking about one or two individual cars at a time here. We’re talking about Southcenter Hill at PM rush in sudden heavy rain.

        And thousands of private automobiles at very close proximity, in varied conditions of repair and maintenance- all under the ultimate control of one computer, very likely in Texas. At speeds around sixty.

        I already own a car with a great many computerized components. I wish I could spec out a few less. But I would never permit any of them to take the control of my car away from me for an instant.

        Really wonder: What if I, as a professional driver, suddenly develop a serious disagreement with the way Ground Control Houston is handling my car and the rest of the herd?

        I tend to avoid freeway travel completely in exactly the conditions this “driverless” technology really envisions, namely without choices, trapped, and unable to escape from the consequences of someone else’s mistake.

        The whole idea of massed separate vehicles all under the second to second control of someone besides their operator only confirms my rule of mode choice: At the exact point where I cease to love driving my car, I want a fast exit to a safe place to park it with a transit station attached.


      2. “all under the ultimate control of one computer, very likely in Texas.”

        I really doubt the cars will be remotely controlled from a different state, the current google car is controlled by computers in the trunk. Navigation to the right destination (as opposed to keeping in a lane and avoiding accidents) is different: GPS is definitely depended on a great deal and the maps originally come from a data center somewhere.

        The very first car to encounter an unexpected obstruction or suddenly adverse weather or whatever certainly has a higher risk of being delayed longer or the car not being able to figure out the proper actions, but if there is wireless communication available it could warn every other car considering the same route. The system would get better with a higher density of autonomous cars (and non/semi-autonomous cars with sensors and communications at a similar level), there are more sensors detecting problems sooner.

        I expect that if road conditions (as reported by other vehicles and sensor systems) exceed some threshold of risk then the car would just refuse to drive, or ask the driver to substitute the google provided insurance policy with their own.

      3. Mike, you’re speculating (rather wildly) about something you know absolutely nothing about. I know next to nothing about it except that your idea of what’s being developed isn’t close to reality. One thing that isn’t coming any time soon is a system where the cars are controlled “from Texas”. All the decisions are made by computers on-board each car, and will be for a long time. Communications between vehicles or with “Texas” has lots of problems, security paramount among them.

        Even in a far-off situation where large-scale decisions (routing around congestion, joining into “car-trains”) were controlled “from Texas” the ultimate control would lie with computers on-board. Steering, throttle, and brake control would certainly be on-board because no wireless network has the latency, bandwidth, reliability, or security necessary to handle all the sensor information and issue detailed commands to the components under control. And the ultimate decision of whether to follow “Texas'” big-picture orders would certainly come from inside each individual vehicle. “Texas” can advise vehicles of construction and congestion conditions, the vehicles figure out what to do about it.

        I absolutely agree that we should have a robust, comprehensive, fast, convenient transit system, and that huge wide highways are part of our dystopian present and we should prefer to scale them down for a less dystopian future. Let’s make sure to make a case against a dystopia that’s coming, not a dystopia that isn’t, so that we don’t miss the point entirely.

      4. In bad weather conditions, computers can’t even tell where the road is.

        Google’s “robot cars” have been driven *only* on well-manicured, well-marked, expensively maintained high-grade California roads, in excellent California weather. Not on the sort of roads you find in rural Minnesota.

      5. Now, I’m badmouthing robot cars, but I want to be clear about this: they could easily replace the vast majority of urban and intercity driving. I am, perhaps, more exposed than average to the sort of driving which they CAN’T replace.

        We could easily end up with some sort of two-tier licensing system, where only the people with the special “manual driving license” are allowed out in blizzards and whiteouts and whatnot, and most of them are on the emergency rescue crews.

    4. This isn’t really a compelling argument. We test it, heavily, until it is shown that the computers are clearly better and safer drivers than humans. And yes, that is a very low bar.

  2. That article is quite interesting…and he goes into a very, very interesting discussion of taxbase efforts of HSR:

    If there were a high-speed rail corridor from Boston to Albany, through Worcester, Springfield and Pittsfield, we would expect real estate in those cities to become more valuable as people fed up with Boston rents moved to smaller cities and the countryside, using high speed rail to commute to schools and jobs. This would have the salubrious effect of increasing the tax base for the most vulnerable communities in MA, though it might decrease the tax base in the most densely settled parts of Massachusetts. Then again, lowered density might be a good thing – few people stuck on I-90 or I-93 on their way into Boston on a Monday morning think the city and its suburbs works especially well at current density.

    1. You wouldn’t expect HSR to be used a lot for daily commuting. The fares would probably be too high to permit that. It only works for the author because he does a weekly commute since he only needs to be on campus 3-4 days a week. In order to do that, he needs a place to stay both in Cambridge and in the town where he lives.

      1. It would depend to some extent on the differential between housing costs. There are reports of people using Eurostar to commute into London from the Pas de Calais on a daily or near daily basis.

      2. For a random Tuesday in January that’s a $120 r/t ticket; the earliest you could arrive at St Pancras is 10am and latest you could leave is 7:30pm. If you’re commuting daily that’s about $2500 a month…I have no idea what the housing costs in Calais are (but am familiar with London). It would have to be a pretty substantial difference.

        Of course, there may be available bulk purchase or discount/pass fares available but those aren’t readily visible online. I suppose for certain people or those who work in the office 3 days or less a week that might be doable. I’d certainly consider something like that if we had fast train service to, say, Cle Elum or Leavenworth areas.

      3. Mark, you’re way off-base. Look at airplanes: guided by ATC, but flown autonomously by autopilot or human pilot.

      4. Just being able to get in quickly, on a regular basis, and without to much forward scheduling is enough for people to fulfill many office jobs.

        So you could work from home, and then commute in twice a week, once a week, maybe even once a month on high speed, or even medium speed rail.

      5. Kyle: worth noting, the autopilot on planes is great in normal conditions, but when something weird happens, it just turns off and suddenly the pilot has to know a gazillion things and handle a gazillion controls accurately.

  3. I’m beginning to wonder if the USA can do anything ‘Right’ anymore.
    Healthcare: We pay more per capita than most countries and have some of the poorest metrics to show for it.
    Manufacturing: So much of what we buy is foreign or foreign controlled. Where did capitalism go to to produce and export our creations to the rest of the world? (each month we run a deficit, we sell a piece of our country – $75B last month, for how many decades now?)
    … and the list goes on and on.
    Transportation: We embark on these ‘Let’s catch up to the rest of the world Missions’, only to fall flat on our faces because our political control struggles seem to choke out every sensible idea that comes along. Years ago,I advocated for beginning the process of finding public ROW segments between SEA-PDX to divorce the state from it’s hostage takers, the BNSF. Long, flat segments do exist, but we choose to spend our HSR money on gold plating the freight corridor for our modern day RR Barons. Just look at the screwing ST takes every time they want to add a train. Yet we continue to pour our limited dollars into the private sector railroads, and they in turn continue to screw us at every turn. Coal and Oil trains at $$/train mile? Bring ’em on boys.
    Is it any wonder why everything has to be sunk in a grossly expensive tunnel to get some grade separation with mass transit, when our politicians run from every good fight – like real light rail down Hwy 99 where it belonged, complete with elimination of street parking, median left turns, and rail/car grade separation at intersections. Real TOD would be flourishing by now, in many different spokes of the wheel, instead of posting about the virtues of S.Bellevue P&R some 20 years after ST began planning for it’s version of Light Rail.
    Driver-less cars are nothing more than a gimmick. Some benefit will come of all this, such as ‘Smart Cruise Control’, where following distances can be set, and where freeway geometry permits and roadway/car interface is modified, so we can let go of the wheel and follow the leader. This is still a long way off, and certainly won’t replace humans for the vast majority of all trips in vehicles. If driver less vehicles were so easy to do, why aren’t all rail systems in the world equipped as Vancouver is? Is it the technology, the cost, or political NaySayers?
    Sorry for the post election rant, but I’ve all but given up our nation can do anything right anymore! I’m moving to Rome, where they at least know how to throw a good party.

    1. Our manufacturing is actually pretty good (if not great). The only time we get beat is when someone else beats us with much cheaper labor (and not just because the cost of living is lower, but often because the working standards are much lower). On the other hand, manufacturing employment is low, which is why we don’t notice the rebound. It simply takes a lot fewer people to make the things we used to make.

      As robots improve, I expect this to happen more and more. It still boggles my mind that we can make a robot that drives a car, but can’t make one to do the dishes. But maybe that is just around the corner. If so, then productivity will make a big leap, but employment will take a big drop — imagine two people in a busy McDonald’s, one to take the orders (with the help of voice recognition and touch screen ordering) and the other to manage all of the machines that cook the food.

      1. We have dishwashers that do the dishes; they require some human input and can’t wash all dishes, but they handle the common cases (silverware, normal-shaped glasses, and plates) and let humans do the rest. Humans have to load the machine and put in the soap; if the dishwasher was invented today it would probably only need soap loaded once a month or so.

        Driving is much more standardized than doing the dishes or cooking fast food. The things that challenge humans are mostly things computers are great at (precisely manipulating a machine’s controls; interpreting feedback from a large number of sensors in different locations; attentiveness). A car’s cockpit has evolved to present relevant controls within arm’s reach of the driver, but a machine can coordinate control right where the power is applied, from anywhere.

        Cooking fast food has evolved into a factory-like situation; as in other food production factories, machines are introduced where they can be, and humans do things they’re better at than machines: picking up and manipulating small, not-quite-uniform objects (like vegetables, which humans evolved to handle, or dishes, which are designed to be handled by humans), talking to people, handling cash in variously weird conditions (cash is a series of small objects designed for human use), etc. Humans, having evolved to eat food, can quickly sanity-check whether food looks and smells good; robots may be improving at this. My brother works for a large bakery company and he’s seen the enormous increase in automation in a recently-built factory-bakery in Atlanta compared to one built in Chicago in the mid-20th century. The older factory has machines that don’t monitor their work and can’t account for inconsistency in inputs, requiring constant adjustment by humans; they can’t handle and manipulate dough, so many intermediate tasks must be done by people standing along a factory belt. The newer factory monitors and adjusts for conditions, and requires less intermediate human intervention.

      2. The robotic cashier I use at the grocery store does pretty well, but it does insist on me presenting the money in a very formalized way. And it sometimes miscounts my change. It talks to me too, but in an annoyingly repetetive way. A lot like a human cashier…

      3. @Al — I guess I don’t see that much difference between driving and loading dishes into a dishwasher (or washing them by hand). An eight year old can do the dishes, but an eight year old can’t drive a car. Part of the reason that driving is difficult is that you have to recognize things. You have to recognize the difference between a pothole and a shadow. You have to recognize what the sign up ahead says. Typically speaking, computers have always struggled with recognition. That is why the robotic car is such a stunning development — it flies in the face of what is normally extremely difficult for machines. Getting a machine to be the best player in chess is child’s play in comparison (simply run through all of the possibilities and determine the best path). But until recently, voice recognition was always “a few years off” (like fusion power). Now, of course, it is here. But like you said, the robot that recognizes a carrot and knows how to slice it is not. Or at least, I haven’t heard about it. Maybe it is just as close to becoming reality as the robot that drives a car, but it doesn’t get as much publicity.

        My point is that as machines evolve, they become more capable, and start doing things that were really hard for them to do not too long ago (e. g. voice recognition). The relatively straight forward transformation has already occurred (as you said, fast food is like an assembly line) but the next step is to have machines do even more of that. It will happen, but like every assembly line process, you won’t get rid of people, you will just have a smaller number managing things.

      4. Driving a car is something humans are not all that well adapted to, evolutionarily. The speeds are greater, the surfaces and motions are foreign, and ultimately we’re controlling many aspects of a very complicated machine. It’s taken a lot of engineering and design, both of cars and roads themselves, to get to the point where humans can, once they’re old and responsible enough, operate cars sort of safely (but still causing lots of damage). I’m reminded of the studies showing that kids that walk to school notice more things about their route than kids given rides; the environment of the road is kept clear, simplified, standardized, so that humans can perform this task they aren’t inherently any good at. This standardization has helped humans and makes it easy for machines. Identifying obstacles and potholes and shadows and lane markings may not be an easy problem, but it’s a solvable problem — you can throw different kinds of sensors and computing power at it, and we’ve been doing that for years. Interpreting non-standard road signs, some kinds of markings, humans directing traffic, etc., is harder. But following the guy in front of you is fairly easy.

        Washing dishes involves manipulating small objects that are not organized in any particular way. If dishes were laid out carefully according to decades of engineering practice, and with entire city layouts dictated by this, as roads are, then machines could wash dishes as easily as they could drive cars. But humans, who are evolutionarily pretty well adapted to manipulating small objects (from vegetables to tools), will always be able to wash dishes more easily than they can drive cars.

        A robot that can wash dishes or slice a carrot under conditions convenient for robots might not be much more difficult to create than a robot that can drive a car, but humans are good at washing dishes and slicing carrots, so why bother with that except on the large industrial scale?

      5. Manufacturing unemployment is low? You mean losing SEVEN MILLION plus living wage manufacturing jobs in the Bush era and our manufacturing base is some how back to normal? I don’t think so…

      6. Manufacturers are having trouble finding skilled employees. It’s the unskilled ones that there’s a lot of, and those who have yesteryear’s skills.

      7. Manufacturers will then need to have direct training programs or empower local trade schools. Expecting the “free market” to magically provide trained skilled labor out of thin air is pretty funny.

      8. “Typically speaking, computers have always struggled with recognition. That is why the robotic car is such a stunning development — it flies in the face of what is normally extremely difficult for machines”

        They’re still very, very bad at recognition. Google is putting its cars in ideal circumstances, namely California.

  4. I’m still waiting for my jetpack, flying car, that thing with 2 wheels so you don’t have to walk, etc…

  5. Regarding the right-of-way issue, we have a great place to put HSR in a future urbanist utopia: freeways themselves.

    I see the future of driverless cars as being fleets of cheap, driverless, electric taxis: basically, a car2go you can use when you’re drunk and that you don’t have to park. Such a ubiquitous and affordable taxi service would incentivize more people not to own cars, rather than swap their existing set of family cars for a similar set of driverless cars.

    A combination of cheap, driverless taxis and high-quality (i.e. frequent, direct, and reliable) transit routes serving trips within, between, and to and from, urban centers is the future of urban transit. (Imagine how much money and diesel Metro will save by not driving marginal coverage routes around unwalkable, low-density suburban areas?)

    This is before we get to the possible applications of driverless technology in the buses themselves. For reasons of politics and public perception, replacing bus drivers altogether would be very hard, but imagine if the bus could assist the driver in pulling up to the bus stop and kneeling the bus almost exactly to the edge of the bus stop curb (EMX is trying to do this with a much cruder technology, albeit without much success so far). Things like this would narrow the gap in ride quality and boarding between bus and rail, at vastly lower cost than laying rail in the street.

    If we can make getting around without your own car easy and cheap, both in the city and the suburbs, so that people are only using their own cars for longer inter-city trips, or weekend drives out in the woods, then I think it will make sense to say “Hey, if we take away half this freeway, we can build you a bullet train that will get you between cities faster than you can drive”.

    I guess the short version is that I see solving the urban and suburban car-free mobility problem as a precursor to solving the intercity mobility problem.

    1. I’m not sure how cheap driverless taxis would be. At best, they replace the cost of the cab driver. But how much of the money spent on cabs is actually going to the cab driver? You also have car maintenance, maintaining the system and the need to make a profit. Last time I checked, renting a car (where I do all the driving) is not exactly cheap. Mass transportation will always be a bargain in comparison, if there are enough people in the area to ride it.

      1. Yes, you replace the cost of the driver.

        If you are comparing private ownership to taxis, then you are going to expend the same costs per mile for both. Same vehicle, same cost per mile.

        However, a taxi can be optimized with technology to pick up others along the way (I think I once figured out it would be almost as cheap for 4 or 5 people to take a taxi in from Kent as it does for fully subsized transit…but those figures are slippery and hard to get…still, maybe subsidizing the cost of a taxi driver for some situations might not be so far fetched).

        But to go back to the points of the article…these are policy not technological issues because 300 mph has existed for decades in other countries! And setting up subsided taxi services, is now possible and being done with private car sharing companies, it’s just that I don’t think the private sector can guarantee service coverage for as many places as they should over as wide a time frame without help.

        So, these things — high speed intra-state transport + a form of last mile service to rail using even human operators is a very real possibility. And I think the only reason we don’t have those things is because we are not being given that choice by our social planners (manipulators really).

      2. Wait, the only reason we don’t have European-style HSR is because America has stronger, more influential “social planners” than Europe does that prevents them from being built? That is an impressive contortion.

    2. “fleets of cheap, driverless, electric taxis.”

      Exactly. Like PRT but using the street grid instead of tracks.

  6. Funny that no one in this thread has pointed out the driverless trains in use in Vancouver, Paris and elsewhere.

      1. Yes, but from what I understand we can only automate trains if we can make sure the entire segment of the track they use is 100% grade separated.

      2. Currently.

        If an automated car can maneuver around obstacles on a roadway, can’t a train stop when it detects an obstacle on the track? Seems pretty simple.

      3. I agree with that. Unfortunately, I’m not the Administrator of the FRA or the FTA.

        But I’m glad that transit systems, which are 100% grade-separated at least at traffic crossings, can already be automated with the technology and policy we currently have.

      4. If an automated car can maneuver around obstacles on a roadway, can’t a train stop when it detects an obstacle on the track? Seems pretty simple.

        This has actually been one of the hardest things for driverless cars to get right. If there’s an obstacle in the right of way, driverless cars tend to get confused and just stop until they figure out if it’s a permanent obstacle (which they should plot a route around) or just traffic (which they should wait for).

        At least with a train, there’s no worries about making unusual maneuvers – you stay on the rails, and don’t move unless they are clear.

      5. Technologically, we should be able to automate lines like Central Link much more easily than we can automate cars on the public streets; even mixed-traffic streetcars should be easier. In terms of regulation, politics, and market forces, though, I’m not sure that’s likely.

        I’d call 10% of vehicle fleet “widespread” — though that’s more than the portion of new vehicles sold that are hybrids, so maybe a lower threshold is warranted. Anyway, if you offered me a $100 bet, I’d bet that 10% of US private car fleet (all vehicles small enough to drive with a standard drivers’ license, including taxis and maybe even delivery vans but of course not tractor-trailers or city buses) is automated before 10% of US passenger trains are automated (in both cases, weighted by anything as long as it’s equal — actively used population, mileage, or hours of operation). I’d be glad to be wrong.

  7. You people really hate it when anyone owns anything, don’t you? News flash: Not even Seattle elected a communist to its city council.

    1. There are many reasons I hope Sawant pulls ahead in the late balloting (a statistical possibility, if not likelihood). But the main reason is that it will force you to shut the fuck up.

    2. Capitalism is predicated on the assumption that humans are naturally rational and utterly selfish, constantly calculating the best result for themselves. Two minutes of observation of humanity should thoroughly disabuse you of the first notion. As for the second, there are actually some outcomes that are more efficient if not driven on an individual basis; ownerless, driverless cars provide one of the more obvious examples of such, since fewer vehicles can move the same amount of people if it’s demand-managed.

      But on a more basic level, why is communism not dead after the Soviet Union collapsed, and in fact is now more popular in this country, supposedly the beacon for capitalism, than ever? I posit that it is because it is the natural state of humanity, but what is also the natural state of humanity is living in groups far smaller than what we have to deal with these days, less than 200 in size. We live in a society far too large for this sort of communism to work, but we still have that instinct for it.

      There’s also the fact that the Soviet Union and most other “communist” countries were less about communism in the sense Marx had in mind and more about getting more power for themselves. George Orwell, who wrote books widely seen as anti-Soviet cautionary tales, was himself a communist, and his books are probably best seen as lamenting the broken promises and “meet the old boss, same as the new boss” outcome of the Soviet revolution (most obviously, the fact that Animal Farm is sympathetic to the revolution at first). Whether intentionally or no, Kshama Sawant seems to be a socialist of this kind, though I don’t think that’s why she lost.

      I suspect socialism is actually the opposite direction from where true communists really should be trying to move; the communism that would work in modern society is probably one composed of small, autonomous groups that collectively serve as capitalist corporations. As such, I think part of the reason the Tea Party has managed to take hold is the fact Democrats seem blind to any solution to the problems facing the working classes that don’t involve more government. If the Tea Party had any position about health care other than opposing Obamacare (and blaming “trial lawyers”), or if it even recognized the problem Obamacare tried to solve, it would have a lot more respect.

      (Part of why Democrats think government is always the answer? We’ve gone so far in embracing the rational, selfish view of human nature that we’re blind to anything else. If rational people acting selfishly isn’t the answer, then the answer must be an institution of rational people acting selfishly channeled in such a way as to “fix” the outcomes, which rarely actually does so but which is seemingly better than nothing.)

      1. “the communism that would work in modern society is probably one composed of small, autonomous groups that collectively serve as capitalist corporations. ”

        You’re describing the “fighting co-ops” model, which is actually pretty successful. I’m involved with several co-ops.

  8. “Google’s super-expensive proprietary radar systems are still way too costly for consumer use”

    I think you are referring to the Veludyne Lidar on the roof, the company isn’t owned by Google but it’s possible they’ve bought a stake in it. A new lidar is in the works that will only cost $10k (instead of $40k or $80k- I think they ought to be able to chop that down in a few years if their sales go up by a coule orders of magnitude.

    1. “coule[sic] orders”? Doubtful. Be sure to price in such minor costs as: all that road maintenance (pay property taxes to enjoy a diet rich in carcinogens? Drivers may word a “get well” card, but their actions say: drive. Which matters more?), the health consequences of all that delightful sitting about (back problems, cognitive decline?), loss of jobs through automating away all the drivers (retrain to farm on Facebook, or something equally edifying?), and I suppose one of these years those of sufficient faith will be rewarded with an energy source that will replace the very small contribution from Carbon in keeping all this ticking along–one ritual invocation involves the words “too cheap to meter,” I believe.

      Walking? Too simple, too many orders of magnitude less expensive, too unpalatable to the populace. So crazy cars it is.

  9. I wrote about driverless cars on my California HSR blog over a year ago and reached some similar conclusions:

    I’ll probably write about Ethan’s blog post sometime this week. I do want to take issue with one thing you’re saying though: “We HSR advocates find ourselves on the losing end of a Betamax-vs-VHS debate, where the technologically superior product fails in the marketplace despite being better.”

    I don’t think we have lost, and I definitely don’t think this is about a marketplace. There is huge demand across the country for passenger rail. But governments aren’t able to meet that demand because Tea Party Republicans have enough power in enough places to block that demand. Once that power is broken – and it will be, sooner than we think – then we will be able to start providing the rail infrastructure that people want.

    1. The Betamax vs. VHS war is interesting, but not perhaps for the reasons people think it is. Beta’s ostensible advantage over VHS was a fairly small resolution difference in what was (by modern standards) a pretty awful picture anyway. VHS was cheaper (both the tapes and the VCRs) and allowed recording longer shows. It turned out that consumers really liked to be able to record movies on a single tape more than they cared about the appearance of the movie. By the time Beta addressed the length of recording issue, VHS had improved picture quality. By the time Beta was manifestly superior on a picture quality basis VHS had already won.

      The lesson here isn’t that inferior products win, but that single metric assessments of technical superiority aren’t generally very interesting. VHS won because it gave people what they wanted at a price they could afford. No amount of technical superiority in some dimension about which people barely care is going to get them to spend more money for a product that doesn’t do what they actually want well.

      I don’t honestly know whether this is a lesson that applies to transit, but it doesn’t do us any good to try to generalize from examples of inferior products winning that don’t hold up to even passing scrutiny.

    2. Thanks for the thought Robert. I meant to say “may find ourselves” as a future hypothetical. I’ve edited the sentence for clarity.

  10. I won’t believe computer-driven cars are ready for primetime until someone takes up Jay Lamm’s challenge and enters an autonomous car in a Lemons enduro. The announcement is tongue-in-cheek (AVs must demonstrate safety and on-track performance on par with those of regular LeMons teams, including the ability to ignore yellow flags; stop dead in mid-corner for no reason; never find the racing line; and argue irrationally when called in for penalties), but Jay is dead serious.

    The kind of conditions found on-track during a Lemons event is a true representation of the kind of worst case conditions an autonomous vehicle would find on a public road – other vehicles of wildly varying speed and unpredictable trajectory, no formalized lane markings, aid vehicles & tow trucks, densely packed high-speed traffic, etc. Just completing one of these events would be a true indication that the control software is ready for prime-time. And the series holds at least 2 events a year at Sonoma, right in Google’s backyard.

  11. Would you rather being a passenger on a completely pilotless jet, or a jet with two pilots? And with that, I believe I just won this argument.

    1. I have no idea what argument you just won, but it wouldn’t shock me if in some future round of airline deregulation this choice literally became available to customers. Which airlines will win? The ones with the lowest fares, as they always do.

    2. Given Asiana’s pilots’ recent performance at SFO when the infrastructure required by the plane to land itself was unavailable, I think I’d take the autonomous plane.

    3. I wouldn’t take the pilotless option with current programming, but in ten years or so, I might change my mind. Plus, pilots have passed tough licensing tests and are required to be fully sober and awake; if they were as nominally-licensed as car drivers, that’d be another matter.

    4. A great way for Google to sell the idea of the safety of driverless cars would be to:

      1) Have their programmers monitor and update software on tablet PCs, as pedestrians in the environment they are testing these vehicles.

      2) Be the first company to have ALL of their employees, management and especially their software engineers do all their long distance travel on pilotless airplanes.

      1. I think it very likely that they would be their own first customer and the result would be a lot of these cars on their campuses (where every employee is a pedestrian), and a lot of employees would start commuting in them to increase productivity. Senior employees and management might even get them first, and it would be a status symbol to have regular access to one.

        The street view cars will probably become autonomous (and passenger-less) as early as they legally can, and this would provide a lot of exposure for the cars outside of the more regularly traveled routes.

      2. I don’t want Google employees to own, nor ride in their driverless cars. In fact, they should be the last people allowed to own one.

        If they trust their vast engineering/programming skills, then they should have no problem being on the outside of 2 tons of metal when they step off the curb in front of one of their driverless cars while engrossed in monitoring the performance of another vehicle they have placed their family in for a driverless trip to the grocery store.

      3. I await the driverless cars which (a) stop and block traffic, and (b) drive off of cliffs.

        Google has not tested its robot cars on anything but well-marked roads. I consider its “300,000 miles” of testing to be a complete joke, worthy of nothing. Run the car through Minnesota dirt roads in a blizzard and I might start taking it seriously.

  12. Pingback: Oh Blog

Comments are closed.