The IT revolution comes to transport

One of the striking features of technological progress over the past fifty years or so has been that of incredibly rapid progress in information and communications technology, combined with near-stasis in most other sectors. Here’s what I wrote on the topic in 2003, and could have reposted, essentially unchanged, a decade later

On most of the obvious measures, technological progress in transport stopped sometime in the late 1960s and, at the frontiers, we are now seeing retrogression.

In 1970, we had regular visits to the moon, and supersonic passenger flight via Concorde was on the way. Now we have neither. Even the space shuttle, designed as a low-cost “space truck” to replace the expensive moon program, is now headed for oblivion, with no obvious replacement.

At a more prosaic level, the 747 jumbo jet, introduced in the late 1960s, is still the workhorse of passenger air transport. Boeing’s attempts at producing a new generation of passenger planes have failed, and the likely replacement for the jumbo jet is the Airbus A380 – essentially just a double-decker jumbo. In all probability, this will be the standard for the next thirty or even fifty years. Of course we don’t have flying cars, or even personal helicopters, as most projections from 50 years ago supposed.

Quite suddenly, this looks out of date. Electric cars, drones and, most significantly, self-driving vehicles have been transformed from curiosities (or, in the case of drones, military hardware with no apparent positive value to humanity) to the likely transport technologies of the near future.

There have been quite a few thinkpieces about these topics, particularly self-driving vehicles, but nothing I’ve seen has been really satisfactory to me. The central focus has been on the challenge of introducing imperfect self-driving vehicles to our current road network. But if we’ve learned anything from the last fifty years (from electronic watches to desktop publishing to digital cameras) it’s that, whatever the initial limitations, a technology that’s been digitised will inevitably improve to the point where it outperforms the analog competition on just about every dimension.

So, it’s safe to predict that, quite soon, self-driving vehicles will be safer and more reliable than all but the best human drivers, and cheaper than vehicles designed for human control. That raises some obvious questions

* what will vehicles be like once the design constraints imposed by the need for a human driver are no longer relevant;

and, more importantly,

* if unskilled or careless human drivers are more dangerous to fellow road users and pedestrians than self-driving vehicles, should they be allowed to drive at all?

To spell out the second point a bit further, if self-driving vehicles are readily available and affordable (and particularly if self-driving technology can be retrofitted to existing vehicles), there’s no argument from the necessity of personal mobility to give speeders and drunk drivers multiple chances to kill other road users. The fact that these people might enjoy driving themselves is scarcely relevant. In fact, to the extent that enjoyable driving is dangerous driving (and, in my limited experience, it mostly is), it’s an argument against allowing it.

Just writing this, I can imagine the ferocity of the responses. I suspect that policies on self-driving cars will turn out to be a long-running front in the endless culture wars in which we seem to be permanently enmeshed.

There’s a lot more to think about here, but that’s enough to be going on with.

84 thoughts on “The IT revolution comes to transport

  1. @derrida derider
    I doubt that safety will be a big factor in the move to driverless vehicles. Safety is generally imposed top-down by the organisations that bear the costs because individuals don’t think very clearly about small risks. There might be a safety-inspired push to get the last hold-outs off the road towards the end as road death becomes denormalised.

    I imagine cost and amenity will be the drivers of change. Many people like driving but we might find other things to do during commute time. We haven’t yet had anyone killed while having sex in a driverless car but that will come.

    Which brings me back to one of JQ’s original post questions which as usual everyone ignores. I think the main change to the physical design of the car will be more specific purposed design. At present, a typical city car needs to be capable of commuting, small goods transport, and occasional trip to the sticks with varying passenger numbers for each trip. In a fleet hire situation you might choose the vehicle type for the task as you do with a hire car. Self driving cars will also free up general in-car behaviour due to the privacy factor and having no driver to distract. Everyone can be drunk. There would be no limit on use of mobile phones or other connected communications. I can imagine a lot of commuters might use the time to clear phone calls (and eat breakfast?) before getting to the office. It is interesting that while most jurisdictions allow hand-free phone use in cars the evidence says this is minimally safer than holding a phone.

  2. I don’t see why there would be any major legal issues with self driving cars. If my car rolls down a hill with no one in it and damages something, I get blamed even though I wasn’t driving at the time. Same if my horse gets frightened by a particularly dangerous plastic bag and bowls someone over. I am responsible whether I was in charge of my totally autonomous horse at that time or not. It seems to me the main issues have been dealt with long ago.

    Also, human beings don’t solve trolley problems when they are about to hit people, so I don’t see why we would expect an autonomous car to. Human beings rely on simple heuristics when about to have an accident. Autonomous cars will do the same, except they’ll do it less often and they’ll do it better. (If they didn’t, we would be unlikely to allow them on the roads.)

  3. @John Quiggin
    I imagine driverless vehicles responding to their immediate environment via cameras and other sensors, reacting to the movement of other vehicles and to the road infrastructure, and this may be adequate much of the time (they’d have GPS too of course), but if the network is going to be maximally safe and efficient, surely some form of collective control would be inevitable, so that traffic flow could respond to conditions a long way further ahead (rain, ice, flooding, crashes, technological glitches etc.

    But I could be wrong – perhaps vehicles could respond autonomously, but even then they would have to have a network of some kind (in addition to GPS) in order to have the necessary information to react to conditions ahead.

    Re the B-double scenario – a human driver will presumably put his own survival ahead of that of other people. I’m wondering about the legal aspects of having an algorithm choose which passenger vehicle to collide with. And I’d be interested to hear from a software engineer how they’d deal with this conundrum.

    But my main concern is the social implications. The dazzling “gee-whizzery” of driverless cars, flying cars etc too often blinds us to the possibility of negative social impacts.

    As far back as the early 60s we were hearing that mechanization and automation would free us of menial tasks and liberate citizens to devote more energy to worthy pursuits (education, the liberal arts, sport, conservation, travel), and of course even the very poor can now have all the electronic entertainments they want and usually a family car.

    But most of the benefits of mechanization have gone to make the wealthy wealthier, while unemployment is stubbornly high, and the coming revolution in automation and robotics will only make things worse, as the gulf between the poor and the rich deepens.

    Does our society have the wit and the will to ensure that the distribution of wealth is fair, that the people who the economy has no use for are able to expect a decent standard of living? Will they be free of the condemnation as bludgers, used even now to justify the meagre allowances they receive?

    I fear a Blade Runner future, with technological luxuries like driverless cars limited to wealthy elites in walled enclaves, while the masses subsist in crowded ghettos and have access only to poorly maintained public transport.

  4. Phil Koopman’s Better Embedded System SW blog:tagged: self-driving cars
    But, while self-driving cars promise improved safety, some critical issues must be addressed to ensure they are at least as safe as human-driven cars and, hopefully, even safer.

    Machine learning is the secret sauce for the recent success of self-driving cars and many other amazing computer capabilities. With machine learning, the computer isn’t programmed by humans to follow a fixed recipe as it was in traditional systems. Rather, the computer in effect teaches itself how to distinguish objects such as a bush from a person about to cross a street by learning from example pictures that label bushes and people.

    Statistically speaking, even 100 million miles of driving is not enough to show that these cars are even as safe as an average human driver. The Department of Transportation should address this fundamental problem head on and require rigorous evidence that machine-learning results are sufficiently safe and robust.

    A key element of ensuring safety in machinery, except for cars, is independence between the designers and the folks who ensure safety. In airplanes, trains, medical devices and even UL-listed appliances, ensuring safety requires an independent examination of software design. This helps keep things safe even when management feels pressure to ship a product before it’s really ready.

    Probably the Defence Trade Controls Act (DTCA) would make significant research in Australia difficult.

    Would inroads into controlling heavy mining/construction machinery be the likely near term usage?

  5. “Statistically speaking, even 100 million miles of driving is not enough to show that these cars are even as safe as an average human driver.”

    I think this is a mistake. It’s true that, in the US, the current death rate is just over 1 per 100 million miles, so, looking at deaths alone, it would be impossible to make a statistically reliable determination. But the crash rate is 185 and the injury rate is 74. That’s enough to make a highly reliable comparison.

    Google’s record so far is 1 minor crash (attributed to machine error) in 1 million miles, which would be incredibly unlikely if the true average rate were over 100.

  6. See How Safe Is The Tesla Autopilot? where Koopman links to an MTBF calculator.

    Looking at it another way, given the current data (1 mishap in 130 million miles), this tool: tells us that Tesla has only demonstrated an MTBF of 27.4M miles or better at 95% confidence at this point, which is less than a third of the way to break-even with a human.
    (Please note, I did NOT say that they are only a third as safe as a human. What I am saying is that the data available only supports a claim of about 29.1% as good. Beyond that, the jury is still out.)

  7. A minor observation re technology. I suggest self-driving cars will have at least 2 driving modes. One for for suburban where it will act singularly and a swarm mode for high volume freeway traffic. The latter probably not centrally controlled rather mimicking schools of fish or birds acting in concert. The militaries are working on these technologies for drones so there is a lot of money being thrown at making it work. I could see the latter having a major affect on congestion.

    Social / economic is more interesting. If you don’t drive, will you still see it as yours or more like a bus where communal use is the norm. Will they remain status symbols. Taxi driving will evaporate taking out one of the starter jobs for migrants. Probably strife from the Shooter & Fisher types defending their rights especially if self-driving proves much safer and Road Safety experts start pushing to ban human drivers.

  8. @drsmithy
    Mission creep! A level of interconnectivity will be irresistible. Enhanced safety may be touted as justification for central control in an emergency, and as the system evolves may be used more routinely. I realize that this is an extreme scenario, but looking ahead 100 years, who knows, maybe not so unrealistic. And I think there is a tendency to be seduced by the exciting new technology and to consequently neglect social factors.

  9. Is there any override ability for the human occupants?

    If not, then that’s scary.

    If so, then the whole thing is a stupid waste – especially of time.

    Free public transport would be far better (coupled with an efficient car-pooling/hire scheme for specialised trips like going outback etc.), and it is achievable – literally – today.

  10. @Ron E Joggles : we have essentially always had centralized control in driving, what’s the issue?

    My freedom to drive down the sidewalk has been impinged upon by law, and my freedom to drive across my neighbors’ properties is barred by the central city planners’ decision not to put a road there (and my neighbours’ decision to place buildings).

    We listen to the government weather reports and corporate media traffic reports.

    We read road signs placed by the government.

  11. @D : human occupant overrides to immobilize the car, almost certainly. Overrides to take a different path, that already exists. I don’t see how those invalidate anything — you just need an extra button for the kill switch.

    The first autonomous cars had steering wheels and pedals, but such fine-grained control won’t be necessary for much longer.

    Personally I don’t miss setting the fuel/air mixture. Manually controlling my gear ratio isn’t that important. Manually pumping my brakes in slippery conditions is passé, thank goodness. I never had the opportunity to yank a chain or turn a crank to start my car. All these tasks have been automated over time.

  12. @Ron E Joggles

    I don’t think there’s any mission creep. Every long-term vision of autonomous cars I’ve ever read about has involved the vehicles interacting with each other over short distances to exchange basic information (“hey guys, I need to take an exit on the left in 4.3km, can you please open up a gap for me in 4.1km”, “a cow is wandering around on the side of the road 500m ahead, everybody slow down”, etc).

    That is a far cry from a single central computer telling every individual vehicle what to do, which is what you seem to be envisaging.

    100 years down the track is far too long to make anything other than wildly speculative suggestions, and if the urge to centrally control every vehicle comes, it won’t be because of the lure of shiny technology, it’ll be because of the lure of authoritarianism (much like people today who want to put a GPS tracker in every car).

  13. @D

    Making public transport free is unlikely to make people who don’t use it today for reasons other than cost (most, since it’s already cheaper than driving if that’s the only metric you use) change their minds.

    Autonomous cars will require manual overrides for many decades. But that doesn’t mean they won’t be quite capable of completely automatically handling the majority of driving scenarios without puny humans taking over.

  14. @Ronald Brakels
    If automated/driverless vehicles become the predominant form of road transport, this will result in a major shift in the regulation of road safety, from the present situation in which the principal legal responsibility for road safety rests with drivers, to a situation in which primary responsibility will rest with vehicle manufacturers and maintenance service providers. This change could potentially occur without any legislative developments since, as you suggest, no new legal concepts need to be developed. I’d surmise, though, that the scope of the potential change would lead to reassessment of the existing legal regime.

  15. @rdb

    To restate my point, the problem is that looking only at fatal crashes throws away 99 per cent of the data. It makes much more sense to look at all crashes or all crashes causing injury. The evidence from Google is overwhelming.

  16. I suppose there might have to be one or two modifications to the law. After all, it would be a drag if we had to wait until our self-driving cars were 17 before they were old enough to get a license. And the procedure for the random breath testing of drivers might need to be modified.

  17. “if the urge to centrally control every vehicle comes, it won’t be because of the lure of shiny technology, it’ll be because of the lure of authoritarianism” – dr smithy @63

    Of course – but then I know which lure is the strongest one. Count on it, the powers that be will eventually have all vehicles centrally controlled. It’s a further move towards “that which is not compulsory is forbidden”.

  18. And let’s not forget all the stuff sent electronically which used to be sent physically. That’s transport too: from mail to music and movies to medical imaging and so on, a lot carrying and couriering has now been made obsolete.

  19. @rdb

    I read the Rand report by Kalra and Paddock, and the reporting of results is highly selective, making the error I suggested and then some. If you go to the linked PDF, and look at Figure 3, you’ll find that, focusing on crashes rather than fatalities, and assuming that self-driving cars are 90 per cent better than humans, it would only take 100 000 miles to prove superiority with 90 per cent confidence. Looking at the actual Google data (1 at-fault crash in a million miles compared to over 100 for humans), that’s very conservative.

    I will write to the authors and see what they have to say.

  20. Auto drive vehicles should drive down the cost of public transport. They should be able to manage spacing between services far better than human drivers can. I’d see first implementation being on our rail ( heavy and light) networks as many of the variables can be eliminated. No overtime rates, we can run public transport 24/7 with reduced overheads

    I wouldn’t want to own a vehicle, an auto drive version of go get would let me call for a car to pick me up, deliver me to wherever and pick up from near my drop off point. It could find its own recharge point in between passengers. I’d pay more for the externalities that are greater than public transport.

    Freight shipping with smart vehicles should be a doddle, avoiding peak congestion both on the roads and at loading and unloading points.

  21. Will our auto drive vehicles automatically take themselves to the mechanics/ service centres for maintenance and updates?

  22. @John Quiggin
    Tesla have built up a very much larger mileage with Autopilot than Google has with its more measured approach. Musk took a hair-raising chance, making a beta system operational on a large scale and relying on feedback to improve it quickly. He got away with it, with only one reported fatal crash (with an irresponsible dead driver who could reasonably be blamed). Autopilot is a partial system, unlike Google’s. But it is heading rapidly to the same place.

    Ikonoclast has the true believer’s faith in the ability of greedy capitalists to escape Prisoners’ Dilemmas by coordination. If the car makers were capable of this, they would have killed self-driving technology. As they all know, self-driving plus ridesharing will probably slash the number of cars needed by an order of magnitude. Instead, they are defecting from the optimum collective strategy and investing heavily. Only a few will survive.

  23. @James Wimberley

    The established capitalists have delayed electric cars and delayed the phase out of coal, by a couple a decades, at least, in each case. They don’t appear to have delayed self-driving technology yet so far as I can tell. This is consistent with the argument I advanced above. Established capital will delay innovative capital’s efforts as long as it can. Eventually, however some innovators become big enough (witness Google and Tesla) to challenge established capital. They do it via new fields of endeavour (mostly) where established capital is not established. Google and Tesla got under the radar of the traditional car makers.

    To sum up, established capital usually can’t kill innovation where it threatens established methods of manufacture and profit but it can delay it. In the end game, those that can do so “defect” and try to survive. Cartel behavior eventually disintegrates under a new challenge of intense competition from a new innovative corporation. With corporations at the mature stage, this is a battle of behemoths, not a battle of tiny, agile entrepreneurs. (Google and Tesla ceased to be the latter some time ago.) The battle of behemoths plays out over decades not years.

    The issue here is that corporate capital (as a system of both “carteling” and competing corporations) probably exhibits characteristics of punctuated equilibrium. There will be periods of relative stasis in some fields at the macro level. The stubborn persistence with internal combustion autos is an example. Finally, new technologies break through and challenge. But the transition is too late from the point of view of saving us from serious environmental consequences.

    Footnote: Where established corporations take up innovations is at the level where it reduces their costs of manufacture. Where they do not take it up is where it would obsolete much of their manufacturing chain (and products) and strand their investment. In that scenario they fight innovation.

  24. derrida derider :
    “Count on it, the powers that be will eventually have all vehicles centrally controlled. It’s a further move towards “that which is not compulsory is forbidden”.


    I wrote at #59, “And I think there is a tendency to be seduced by the exciting new technology and to consequently neglect social factors.”

    Almost every other commenter here discusses only the technical detail, without consideration of the social impacts of ever more pervasive interconnectivity, and the inescapable potential for surveillance, influence and ultimately, control.

  25. My god! Centrally controlled autonomous cars! Do you realize what this means? To control literally millions of vehicles centrally, Australia would actually have to have working, reliable, high speed internet!

    I am so stoked!

    NBN via a boot stomping on a human face forever!

  26. @Ron E Joggles

    I’m all for looking at the social factors, but I’m still at a loss to work out what your concern is. Is there anything that could be done with the (not currently feasible) idea of remotely controlling autonomous cars that isn’t already being done for commercial drivers, much more cheaply, with GPS tracking. Drivers can be ordered to follow travel plans and punished for deviations.

    Relative to either of these, I’m much more concerned at the likelihood that this is already happening to the general public, surreptitiously, using mobile phones as the tracking device.

    And, of course, it’s not as though driving has ever had an assumption of privacy – it’s an easily observable activity, undertaken in public.

  27. Cars won’t be centrally controlled, because as any 14 year old can tell you, lag gets you killed. And then after you’re killed you get a heap of American accented racist and sexist verbal abuse. I always assumed the verbal abusers were also 14, but as it turns out, they were apparently old enough to vote.

    Cars won’t be centrally controlled, but if the technology did work that way I would be fine with it as long as it saved lives.

    What would worry me is if the government started doing things that served no practical purpose but only had evil applications. For example, if they started building concentration camps in remote areas that Jews, Gypsies, or other groups that could be classified as “enemies of the state” or “un-Australian” could be imprisoned in without trial. If they blocked media access to these camps. If they built up a workforce of camp guards and administrators who could be relied upon to perform inhumane actions when told to by those in power. If they inflicted physical harm on children, pregnant women, the sick, the elderly and other groups that society normally regards as requiring special protection and got away with it.

    That’s what would get me worried.

  28. I’ll be seriously worried once they start censoring the XXXXXXXX XXX XXXXXX XXXX XXX XXXXXXX XXXXXXX XXX XXXXX X XXXXX XXX XX XXXXXX XXXXX XXXXXXX

  29. On the subject of:

    * what will vehicles be like once the design constraints imposed by the need for a human driver are no longer relevant;

    Well, the elimination of controls means there is space for an extra seat in the front. This means self driving taxis of similar interior size to current ones could seat 6 passengers instead of 4.

    Of course, self driving cars could be made narrower than current ones so more can fit side to side on the roads. Once human driving is eliminated permanent lanes won’t be necessary.

    Self driving cars will electric. Batteries can go in the floor to give a low center of gravity and an electric motor can go in each wheel. This gives a lot of freedom with what goes on top of the “skateboard”.

    There might be wide variation in car size. As most taxi trips consist of ferrying on passenger, perhaps the majority of vehicles will be single seat self driving taxis. Or maybe private citizens will feel the need to express themselves with giant land yachts. (Donald Trumps preserved head will have its own self-driving continental siege machine that goes by the name “Mark 4”.)

    One development we might see very soon is external airbags. Self driving cars would (hopefully) be able to deploy them just before impact.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s